key: cord-0808732-sfve0684
authors: nan
title: EPSM 2020, Engineering and Physical Sciences in Medicine
date: 2021-08-23
journal: Phys Eng Sci Med
DOI: 10.1007/s13246-021-01024-z
sha: c5de8588e51d7312af748f8bb8573e4bd3cd8462
doc_id: 808732
cord_uid: sfve0684

nan

Introduction The Australian Institute of Physics (AIP) recently released a position statement regarding replacement of face-to-face (F2F) classes by online delivery in physics courses impacted by the COVID-19 pandemic [1] . It concluded online teaching should only be viewed as a short-term measure during an emergency situation. The arguments presented against online teaching were related to preserving the excellent reputation of Australian physics teaching which is based on a) the largely F2F teaching methods and hands-on curriculum b) high levels of student-student and teacher-student interaction and c) invigilated examinations. The seven course coordinators from Australia and New Zealand's (ANZCCs) ACPSEM accredited postgraduate programs considered the relevance of these arguments to medical physics education. ANZ Medical Physics Courses All six universities offering coursework (rather than research)-based degrees transitioned from F2F teaching to online in a period of a few days to \2 weeks. The ANZCCs held monthly meetings to provide mutual support. Most ANZCCs reported good, if not better attendance at online lectures compared with traditional F2F lectures. Engagement in lectures Introduction The COVID-19 pandemic has impacted Radiation Oncology clinical departments in a variety of ways, each of these having a flow-on effect to Radiation Oncology Medical Physics (ROMP) training, education and assessment program (TEAP) registrars. The ACPSEM made the decision to put measures in place to continue the external assessment of registrars without compromising the quality of the assessment, the safety of the registrar and examiner (and their respective departments) and abiding by various state/national COVID-19 restrictions. Method An alternative method for Part A and B examinations needed to allow for examiner/registrar interaction and visualisation of practical equipment setup as well as supervision of the exam to ensure appropriate examination techniques were followed. Zoom video conferencing was used for Part A and B examinations with two examiners remotely assessing the registrar with a third examiner (local to the department) on-site to ensure exam processes were followed without being involved in the grading of the candidate. There were also concessions for the written examination. Candidates not able to complete the brachytherapy competencies due to COVID-19 restrictions were given the opportunity to undertake the other core module written exam questions, with the brachytherapy module being completed at a later date. Results To date, 15 registrars have been able to successfully complete TEAP final exams despite COVID-19 restrictions. Post-exam evaluations provided by both examiners and registrars indicated that the Zoom examination method was fair and the examination standard was maintained.

To date, 1 registrar has undertaken the written exam without having completed the required brachytherapy competencies. This registrar is hoping that restrictions will ease and allow them to finish the outstanding brachytherapy competencies so that they can complete the brachytherapy written exam questions in March 2021.

Conclusion Despite the challenges COVID-19 has placed on TEAP registrars, the ACPSEM has adjusted its examination methodologies to continue to allow registrars to progress through the program. References/Acknowledgements The authors acknowledge the flexibility of the examiners, candidates and their departments, who were prepared to use these alternative methods.

O003 Simulation and problem-based learning: engaging medical physics students in active and authentic learning Introduction We report the results of a project to investigate if replacing traditional lectures in a radiotherapy unit with simulation and problem-solving activities could increase student engagement and enhance student knowledge and understanding of complex radiotherapy physics concepts. Method The first of two new activities involved the students using a BEAMnrc/DOSXYZnrc Monte-Carlo model of a linear accelerator (LINAC) and water phantom. The activity aimed to give students hands on experience of the production of x-rays and the role of the different components of the LINAC on the clinical beam as well the resulting dosimetric data. The second activity involved using the matRAD inverse treatment planning software to perform comparative planning of protons vs photon IMRT two clinical cases. The aim was to give the students hands on experience of inverse planning optimisation as well as an understanding of the potential advantages of protons over photons. A qualitative mixed methods approach was used to explore student experiences with the unit. Class observations, student feedback through anonymous online surveys, focus groups and staff reflections were used to gain an in-depth understanding of the way in which students engaged with the new activities in the unit. Results Overall the results indicate that the approach was well received by the students who valued the opportunity to use sophisticated software to simulate two different aspects of clinical radiation therapy. Although there were obvious challenges associated with using potentially resource-hungry software on students' standard laptops, there was an observed improvement in student engagement with subject content as well as engagement with the instructor and between peers. Conclusion Replacing traditional face to face lectures with clinically relevant hands on simulation and problem-solving activities in a postgraduate medical physics course was found to increase student engagement with subject content as well as engagement with instructors and between peers.

O004 The role of Diagnostic Imaging Medical Physicists in supporting clinical policy change and the implementation of evidence-based clinical practice of Medical Imaging and Nuclear Medicine, Queensland Children's Hospital, Brisbane, Australia (kerrie.norynberg@health.qld.gov.au)

Introduction Traditionally medical physics service support within Diagnostic Imaging departments has been thought of as the provision of radiation safety services and fulfilling regulatory requirements concerning equipment compliance. The aim of this paper is to illustrate, using a clinical example of the cessation of the use of paediatric gonadal shielding, that the implementation of a major clinical policy change requires medical physics input at all stages through the process. Following the publication in the literature of a number of position statements supporting the termination of the use of gonadal shielding in paediatric pelvic imaging there was a substantial body of work required to implement a departmental policy change. Medical physicists were needed in order to ensure scientific rigour and to provide essential scientific support. Method Evidence and examples of scientific input from medical physics staff was provided in all aspects of changing the use of gonadal shielding within the department: 1. Providing a background review of the literature and current evidence 2. A review of current practice within the department 3. Consultation with radiologists and radiographers 3. updating the Radiation Safety and Protection Plan 4. Provision of education to all staff involved in imaging of paediatric patients 5. Development of support materials to be provided to patients/parents/carers to explain the reasoning behind the policy change 6. Formulation of a research protocol (including ethics submission and approval) to audit the process 7. Analysis of results 8. Communication to the wider imaging community.

This reflection has shown that through this process the contribution of Diagnostic Imaging Medical Physicists is invaluable in ensuring that medical imaging services are operating to the highest standard of evidence-based practice. Sharing this example shows the impact of a DIMP service to the clinical department.

O005 Assessment of adaptive radiotherapy workflows for head and neck cancer Introduction Inter-fractional anatomical variations in head and neck cancer patients can lead to clinically significant dosimetric changes. Adaptive re-planning should thus commence in order to negate any potential over-dosage to organs-at-risk (OAR), as well as potential under-dosage to target lesions. The aim of this study is to explore the correlation between transit fluence, as measured at an electronic portal imaging device (EPID), and DVH metrics to target and OAR structures. This work will guide implementation of transit EPID dosimetry as a decision support metric for head and neck adaptive radiotherapy. Method Ethics approval has been obtained to conduct a retrospective analysis of 20 patients who have completed a course of radiotherapy at the RAH with at least one re-plan. The purpose of the study is to investigate the correlation between change in transit radiation fluence and change in key internal dosimetric parameters. An in-house developed RayStation script will allow for transit fluence simulation on planning CTs. The RayStation script was validated by comparison to EPID measured transit dosimetry fields using a phantom on a TrueBeam linac. A modified gamma analysis tool was developed to quantify changes in transit fluence with changes in phantom geometry.

Results A comparison of linac measured change in transit fluence and RayStation calculated change in transit fluence, when a solid water phantom is reduced in thickness (1 cm removed anteriorly and posteriorly), is shown in Fig. 1 . A gamma analysis was performed on images acquired with a full phantom and a reduced phantom thickness. Figure 1 indicates that the change in RayStation transit fluence reflects change in true transit fluence as measured on a linac. Conclusion The RayStation transit dosimetry script developed was validated against measured data. The tool can thus be utilised to explore the correlation between transit fluence and DVH metrics to target and OAR structures in future work.

O006 Timely review of electronic portal images in a multicentre clinical trial of deep inhalation breath hold in breast cancer patients Introduction Deep inhalation breath hold (DIBH) is considered a good method to reduce heart dose in left-sided breast cancer patients undergoing radiotherapy. The HART trial (TROG 14.04) aimed at assessing this in a multicentre setting with one of the endpoints being reproducibility of breath hold position. Method The trial accrued 32 patients with left sided breast cancer in six contributing centres in Australasia. DIBH was controlled using Varian RPM or Elekta ABC devices. Reproducibility of breath hold was assessed during the first week of treatment using daily electronic portal imaging (EPI). The images were uploaded to the central quality management system of the TransTasman Radiation Oncology Group (TROG) and reviewed by an independent physicists using mid lung distance (MLD) as a surrogate for patient positioning. Reviews were conducted using MIM software and many of the images were reviewed again after completion of the trial by an independent observer together with additional images acquired over the remainder of the treatment. Results Two of the 32 left sided breast patients accrued did not proceed to DIBH (one clinician and one patient decision). Image review for the remaining 30 patients was completed within three or less working days after the first week of treatment and communicated to the treating centre. Three images of 177 were found to be of insufficient quality for review. Inter-observer reproducibility was good (mean r2 = 0.734, n = 10 image sets with 5 or more dual observations, Figure A) . Figure B shows as comparison the MLD judged by the same observer from medial and lateral images provided for six patients (mean r2 = 0.469). Conclusion Review of multiple verification images acquired during a clinical trial is feasible and provides a feedback or intervention point taking actual treatment delivery and not just the treatment plan into consideration.

Introduction Adaptive radiotherapy is complex, with multi-disciplinary processes and the potential for multiple deformable registrations reinforcing the need for patient specific quality assurance [1] [2] [3] [4] . A practical and robust method for assessment, evaluation and reporting of offline adaptive cases is ideal. This study aims to characterise a standardised solution developed by the ACPSEM Medical Image Registration Special Interest Group (MIRSIG). Method A working group was created based on an expression of interest amongst MIRSIG for the development of a quality assurance reporting template compliant with AAPM TG132 [4] recommendations for Australian and New Zealand radiotherapy sites. The template was developed by initial prototype, alpha testing of branched versions, survey of working group, and development of the consensus template with core ([75% agreement) and optional checks. Results 16 survey responses were analysed to generate a consensus template to facilitate multi-disciplinary tasks and handover for Radiotherapists, Physicists, and Oncologists. Core checks identified were divided into 4 major processes ( Figure 1 ):

• Request (Radiotherapists): Date of Request (88%), Patient Identifier (94%), Rationale for Request (88%), Fractionation (delivered/total) (88%)

Introduction ARPANSA's Diagnostic Reference Level (DRL) program supports optimisation of medical exposures by providing guidance on typical dose for common imaging procedures. Routine comparison against DRLs is a component of quality assurance and is included in regulatory requirements such as the Medical Exposure Code [1] and the Commonwealth Diagnostic Imaging Accreditation Scheme [2] . Method ARPANSA's National Diagnostic Reference Level Service (NDRLS) collects CT survey data through a web portal. Data on image-guided interventional procedures (IGIP) is collected by e-mail using spreadsheet templates. Median values of dose metrics from each survey are reported as facility reference levels (FRLs). Surveys include data for up to 20 patients in CT and 30 patients for IGIP. National DRLs are based on the third quartile of the distribution of FRLs.

Results In 2019, 4781 CT surveys from 683 scanners were completed. Third quartiles of the FRL distributions for each dose metric, categorised by scan region, are shown in Table 1 and compared with the national DRLs. The data received are consistent with the national DRLs. There is now sufficient data to establish a DRL for KUB scans. For IGIP, 47 surveys were completed in 2019. A liaison panel reviewed the data collected through to the end of 2018 and recommended the following national DRLs for diagnostic coronary angiography: dose-area product (DAP), 30 Gy.cm 2 , and cumulative air kerma at the reference point, 0.5 Gy. These DRLs were published on the ARPANSA website on 31 March 2020 [3] . The panel also made recommendations to revise the procedures included in the IGIP survey. Conclusion Australia's DRL program is now well established. Current data for adult CT is consistent with the national DRLs. There is now sufficient data to establish a DRL for KUB scans. National DRLs for diagnostic coronary angiography have been established.

References/Acknowledgements 1. Australian Radiation Protection and Nuclear Safety Agency (2019) Code for Radiation Protection in Medical Exposure. ARPANSA. https://www.arpansa.gov.au/regulation-and-licensing/ regulatory-publications/radiation-protection-series/codes-and-standards/ rpsc-5 Accessed 31 May 2020 2. Australian Government Department of Health (2015). DIAS Practice Accreditation Standards (from 1 January 2016). Department of Health. http://www.health.gov.au/internet/main/publishing.nsf/ Content/di-DIAS-Prac-Accred-Standards-1-January-2016. Accessed 31 May 2020 3. Australian Radiation Protection and Nuclear Safety Agency (2020) Current Australian national diagnostic reference levels for image guided and interventional procedures. https://www.arpansa. gov.au/research-and-expertise/surveys/national-diagnostic-referencelevel-service/current-australian-drls/igipAccessed 31 May 2020

Introduction ARPANSA has developed and published free online Occupational Radiation Exposure material for all people working in medical facilities that use ionising radiation, for example those using X-rays/CT and nuclear medicine. The modules are brief (10-35 minutes in total) filling a 'gap' identified by medical physicists, trainers, RSOs and regulators.

The ORE modules provide useful information on occupational radiation protection and safety for everyone, ranging from staff in local medical imaging facilities through to those in our major hospitals. The material is tailorable by occupation (such as nurses, cleaners, radiologists etc.) and by the level of involvement with radiation (for example, those in administration only need a limited amount of information, compared to staff in a nuclear medicine department).

Method ORE for medical facilities was developed iteratively with key stakeholder input including the professional colleges, regulators, facilities experts and end users, using modern learning strategies and techniques.

Results ORE is an innovative set of training material. Its modular structure and flexible navigation allows for the key medical professions and support functions to individually tailor the content to better suit their needs. The modular nature also makes additions and changes easy. Most facility staff will complete the module as a tailored online course via the ARPANSA website, with no login required. The training material can also be downloaded to an organisation's SCORM compliant eLearning system or as an interactive PDF/Pow-erPoint to facilitate group question and answer sessions.

Conclusion Evidence has shown that all staff in medical facilities utilising ionising radiation can benefit from increased medical radiation safety awareness, and a common language and basis for understanding assists everyone. ARPANSA's ORE training materials fill a much needed gap. Feedback is still being sought, with the improvement process ongoing. Introduction Physicists and medical professionals often rely on commercial dosimetry software as a time-efficient way to estimate radiation dose for radiological examinations. In many cases, there are significant variations between the organ doses calculated by different dosimetry software, particularly in younger paediatrics. Infants are among the most radiosensitive age cohorts due to their rapidly dividing cells and longer time for possible malignant disease onset. Despite this, no studies have verified the organ doses provided by commercial software in infants undergoing computed tomography (CT) examinations. This study aims to verify the accuracy of organ doses estimated by CT Expo and NCICT for a 1-year old phantom by comparing them to thermo-luminescent dosimeter (TLD) measurements.

Method 85 high-sensitivity LiF (Mg, Cu, P) TLDs were embedded within a CIRS whole-body 704B one-year old anthropomorphic phantom. The phantom underwent vertex-toe CT examinations using three different CT scanners using standard hospital imaging protocols (with increased tube current). The TLDs were calibrated in dose to water using a superficial x-ray radiotherapy unit scans. Additional CT examinations were performed with different kVp and added filtration. Doses were estimated using CT Expo v.2.5 and NCICT v.3.0 using the scan parameters and patient characteristics which best suited the CIRS phantom within each application, respectively. Results The measured data indicates an over-estimation of organ doses calculated by the dosimetry software. Except for the thyroid, NCICT estimated organ doses closer to TLD measurements than CT Expo. Most organ doses estimated by NCICT were within 20% of TLD measurements across multiple kVp and CT scanners. Some CT Expo organ doses varied by more than 50% when compared to TLD measurements. Figure 1 indicate a subset of results for a single tube voltage and scanner.

Conclusion The study indicates dosimetry software tend to overestimate radiation dose in one-year olds with NCICT providing the best estimate for most organs. Introduction The portability of hand-held dental X-ray units makes them valuable for use in aged care, forensics [1] , and schools. However, the close proximity of the operator to the X-ray unit raises radiation safety concerns. Tests were conducted using a Rextar X unit to evaluate the radiation dose to the operator during dental X-ray exposures.

Method Leakage and scattered radiation were measured with an Unfors Xi Survey detector. Scattered radiation was generated via X-ray exposures of a CTDI head phantom. Scatter dose measurements as a function of distance were made with and without a lead acrylic scatter shield (0.6 mmPb at 100 kVp) attached to the unit. Measurement locations were chosen to reflect the position of the operator's hands and eyes.

Results Leakage was well shielded, with a maximum of 0.02 uGy detected at the surface of the Rextar X unit. Without the scatter shield, hand dose to the operator from a typical adult bitewing exposure was 0.69 uGy (left) and 0.79 uGy (right). Introducing the scatter shield reduced these doses by approximately 75%, to 0.25 uGy and 0.12 uGy respectively. Scatter profiles obtained in the horizontal and vertical directions, with and without the scatter shield attached are shown in Figure 1 .

Conclusion The Rextar X is well shielded, giving very little leakage radiation dose to the operator. Even without a scatter shield attached, it would take over 1000 exposures to reach 1 mGy absorbed dose to the hands and eyes. Scattered radiation to the operator is significantly reduced when a scatter shield is attached. Although it is highly unlikely that operators would reach annual dose limits if using the Rextar X without a scatter shield, it is strongly recommended to attach one for compliance with the ALARA principle.

References/Acknowledgements

Hand-Held Dental X-ray Equipment. Public Health England. https://www.gov.uk/government/publications/hand-held-dentalx-ray-equipment-guidance-on-safe-use. Accessed 18 May 2020 I would like to acknowledge Thomas Greig for his advice for this project. I also acknowledge Karen Sleishman and the team at Newcastle Dental Clinic for providing the Rextar X.

Introduction Intrafraction imaging allows visualisation of tumour or surrogate during radiation therapy. As the dose from kV imaging is difficult to measure directly, Monte Carlo (MC) simulations can be used to obtain estimates. The imaging dose from intrafraction kilovoltage imaging was simulated for different stereotactic ablative body radiotherapy (SABR) treatments. The dose increase to target volume is presented.

Method MC software (ImpactMC) was benchmarked to match beam quality and output for the on-board imaging system of a Varian Truebeam linear accelerator. Dose/frame was calibrated by simulating a cylindrical water phantom (radius = 15.3 cm), using parameters from [1] . CT scans of prostate, liver, lung, spine, and pancreas patients were used as input for simulation. Imaging system was characterised with the following parameters: 125 kVp, Ti filter (HVL = 7.98 mm Al); 80 mA, 13 mS, 7.03 mGy/100 mAs. Field size, imaging frequencies and fractionation listed in Table 1 Introduction The ICRU has produced guidelines for how to report and prescribe 3DCRT and IMRT [1-[3] . In its radiotherapy clinical protocols, our centre states how it complies with these guidelines. For 3DCRT sites, we prescribe to the ICRU Reference Point, and for VMAT/IMRT sites, we aim to prescribe to the target PTV(s) D 50% (we allow head and neck plans prescriptions to vary by ±1.5% and prostate plans to vary by ±1.0% from D 50% ). An audit has been performed on our prescribing practice to verify the compliance with our stated aims. Method Our Eclipse [4] TPS patient database was data-mined using an in-house application developed with the Eclipse Scripting API. The application extracted the plan date, site, prescribed dose, PTV IDs, D 50-% dose, and ICRU Reference Point dose. Data for all sites for all our clinical protocols that stated a prescription to a point or volume were analysed.

The results are shown below for two body sites, head-andneck VMAT treatments (Fig. 1A) , and 3DCRT pelvic treatments e.g. rectum (Fig. 1B) .

Conclusion VMAT -we are a) reviewing our plan optimisation process to better achieving the planning aim and b) looking at finalising prescriptions only after plan optimisation is complete. 3DCRT -During the planning process we modify the plan normalisation to typically achieve a mean dose of 100%-±-1%, then place the reference point on the 100% isodose. This results in plans with a large variation (±-2.4%, 2%) in the ICRU Reference Point dose. However, it is not what we state in our clinical protocols that we do, does not follow ICRU-62 or 83 guidelines, and is not consistently applied. In order to address this situation, we will be moving as soon as is Figure 1 Trends in systematic volume differences in whole heart delineations, defined as automatic -manual (left), and mean whole heart dose (right, shown for left-sided patients receiving 40 Gy/ hypofractionated, or 50 Gy/standard treatment). Thick lines/shading = 1-year rolling median/IQR Phys Eng Sci Med placements of landmark points is very time consuming and not practical. Image matching by using a set of local interest points using the Scale-invariant feature transform (SIFT) tool was proposed by Lowe [2] . This tool can automatically extract dense landmark points. This work investigates a framework to integrate and validate the SIFT tool for use with a clinical system. Method Figure 1 describes the tool's client-server architecture to compute the matching landmark points into existing clinical systems i.e. client (MIM software).

Client-server architecture to integrate the SIFT tool into clinical system

Either entire scan length is used or a region of interest (ROI) is created in MIM. The user runs the tool and landmark points are generated. Plastimatch [3] is used to compute matching SIFT points which are imported back into MIM. This tool was run on over 47 DIR head and neck image datasets which were acquired over different time spans and scanned in different orientations. Eight entire scan length and 44 ROIs over 40 datasets were considered. 1200 resulting landmark points were assessed to verify anatomical correspondence by a radiation oncologist. Results An average of 12±5 landmark points were computed within each ROI and 100±15 over entire scan length. The majority of landmark points were detected around tissue-air or bony interfaces. Only 23 points out of the 1200 landmark points were found to be unacceptable. The time to compute the points within desired ROI takes up to 3 minutes or up to 10 minutes over the entire scan length. Conclusion The SIFT tool was successfully integrated within a clinical system and able to compute dense and reliable landmark points. The time to run this tool was found to be reasonable for clinical use. Introduction This research is aiming to protect bladder and rectum through giving them PRV (planning organ at risk volume) margins. Method For each of the total of 16 patients, the bladders and rectum were delineated on CBCT images in five fractions in addition to the CT image set. Then the bladder and rectum wall displacement were measured through comparing CBCT and CT contours. Based on the measured data, the relationship between the organ wall displacement frequency and percentage distance corresponding to the biggest organ wall displacement was evaluated. According to this relationship, PRV margins could be created to cover a specific percentage of organ wall motion for a specified percentage of the population.

Since the rectum wall displacement decreases from superior to inferior parts. In this research, the rectum was segmented into 3 different parts (Figure-1) based on rectum flexure and midline of the rest part of rectum in GTV. The required PRV margin for each part is suggested separately. Results For the bladder, a PRV margin of 0.75 cm (right), 0.75 cm (left), 1.00 cm (anterior), 0.80 cm (posterior), and 0.55 cm (inferior) could cover at least 90% of bladder wall outward motion (the wall motion caused by organ volume expanding) for 90% (14/15) of patients. If the effective bladder volume controlling method could be applied to control the bladder volume, a smaller bladder PRV margin with a width of 0.60 cm (right), 0.65 cm (left), 1.00 cm (anterior), 0.80 cm (posterior), and 0.45 cm (inferior) could be used to achieve the same goal. For the rectum, a PRV margin as given could cover at least 90% of the rectum wall outward motion for 90% (12/13) of the patients. Table 1 Derived rectum PRV margins for sections of the rectum Conclusion A series of PRV margins were generated to meet the bladder and rectum motion coverage objectives. However, the clinical application of these margins needs comprehensive discussion.

O018 Can we increase PTV margins for lung cancer patients undergoing radiotherapy with deep inhalation breath hold?

Introduction Deep inhalation breath hold (DIBH) is a commonly used method for motion management in radiotherapy. In addition to 'arresting' the motion of the tumour it reduces lung density. We investigated how margins could change if the same mass of lung was irradiated in DIBH and free breathing (FB). Method Lung density was assumed a surrogate for functional lung tissue and assessed using Hounsfield numbers in planning CT scans (Philips Brilliance wide bore) of thirteen breast cancer patients treated in DIBH. This was compared with density during FB. Lung density in four locations ( Figure 1 ) was found to be on average 28 +/-9% lower in the anterior parts of the lung compared to posterior ones. This resulted in more potential for lung sparing in the posterior aspects (49% density reduction compared to 42% in the anterior parts). The average density reduction was used to calculate the increase in margin that would result in the same amount of lung in the high dose region during DIBH as in FB. Different spherical tumour volumes and planning margins were studied. Results The additional margin facilitated by reduced density of lung during DIBH increased with increasing tumour volume, increased original margin size and the density of lung during FB ( Figure 2 ). Margin reduction in three dimensions ranged between 1.7 mm for a typical lesion in stereotactic ablative radiotherapy (SABR) to 7 mm for a large mass with a 15 mm margin. In case of one-dimensional extension (eg in direction of breathing motion) the additional margin can be significantly larger. Conclusion Assuming that lung toxicity is reduced with the density of lung in the high dose volume during radiotherapy, DIBH allows an increase of PTV margins without expected increase in toxicity, thereby potentially accounting for additional uncertainty in the reproducibility of tumour location in DIBH.

O019 Resolution of cH2AX in head and neck cancers following fractionated irradiation: comparison by HPV status Introduction Significantly better responses to therapy in head and neck cancers that result from the human papillomavirus (HPV), compared to other causes, continues to be an important focus of research activity. The positive prognostic status of HPV in head and neck cancers demonstrates a potential for dose de-escalation and a more personalised approach to therapy. A more developed understanding of treatment response, however, is required of these cancers to differentiate treatments between the 2 aetiological groups. This study is presently on-going, investigating HNSCC repair of DNA damage from fractionated X-ray dose, using measures of cH2AX resolution in cell lines following 4 Gy fractions and comparing between HPV positive and negative status. Method Three HPV positive and 3 HPV negative HNSCC cell lines are grown and irradiated with 4 Gy fractions. Re-culturing of cells surviving 4 Gy irradiation forms the next generation of that cell line which is then irradiated with the next 4 Gy fraction. Cells of each generation of each cell line are fixed and permeabilised at 30 minutes and 24 hours following irradiation prior to staining for cH2AX expression and flow cytometry. Results Comparisons of cH2AX recovery between 3 generations of cell line UPCI-SCC-090 show the greatest resolution in the 1st generation, from 60.4% expression at 30 min to 9.3% at 24 h. The 3rd generation (following 3 9 4 Gy fractions) demonstrated the least Figure 1 Locations of CT number measurements in patient. Diameter of the four regions of interest was 25mm Figure 2 Additional margin possible due to reduction of lung density during DIBH. This is considering both 1 and 3 dimensional expansion of the PTV as well as different original margins for a spherical lesion of different sizes Phys Eng Sci Med recovery, 41.7% expression at 30 min to 14.8 at 24 h. Background expression of cH2AX increased from the 1st generation level of 4.6% to 7.6% in the 2nd generation and 14.3% in the 3rd generation ( Fig. 1 ). Conclusion Both a diminishing capacity for cells to achieve cH2AX recovery and an accrual of unresolved cH2AX foci is observed in subsequent generations of the same cell line.

O020 Auto-segmentation of the clinical target volume in a gastric cancer clinical trial Introduction Consistent delineation of the Clinical Target Volume (CTV) can become challenging in clinical trials due to an often-large number of centres involved. This is the case for the international TOPGEAR clinical trial [1] where patients in one treatment arm will receive radiotherapy before surgery and the protocol for delineating the CTV requires certain anatomical landmarks to be included/ excluded from the structure. An auto-segmentation tool could help automatically flag treatment plans where a treating centre has potentially incorrectly defined the CTV. Method Five Radiation Oncologists provided ground truth delineations of the CTV on 10 atlas cases. A consensus workshop was held, where these five clinicians made changes to their delineations as necessary, to ensure that all contours conform to the TOPGEAR clinical trial protocol. A multi atlas-based auto-segmentation approach [2] was used to segment the CTV structure using this atlas set. A leave-one-out analysis was performed to assess the quality of the auto-segmentation approach and the Dice Similarity Coefficient (DSC) was computed to compare the segmentation to the observer's (ground truth) delineations. The mean DSC between each observer was also computed to compare the auto-segmentation quality against inter-observer variability.

Results The mean DSC of the auto-segmentation was 0.8±0.05 compared to 0.89±0.02 for inter-observer variability (Table 1) . For half of the cases the auto-segmentation DSC score was near to what was seen in inter-observer variability (\= 0.05 difference). Conclusion The auto-segmentation performs well in most cases given a relatively small atlas set. The next step will be to use the probability map produced by the multi atlas-based auto-segmentation approach to quantify the uncertainty of the auto-segmentations. This tool could then help provide quality assurance for treatment plans within the TOPGEAR clinical trial as well as future gastric cancer trials. References O021 Deploying auto-segmentation tools in the clinic: a flexible, robust and scalable framework Introduction It is often desirable to make auto-segmentation tools developed within medical physics/radiation oncology research projects available in the clinic. Ideally, these tools would integrate directly into existing workflows and clinical systems. Due to the nature of many research projects, this can often be quite challenging and require significant additional work. The framework presented in this work aims to overcome these challenges and ease the process of making auto-segmentation tools available in the clinic.

Method To ensure the framework is flexible, robust and scalable, a client-server architecture was selected ( Figure 1 ). Like this, the client-side can prepare data and send it to the server for processing and finally download the auto-segmentations. This gives the framework the flexibility to be integrated into numerous systems already being used clinically. By implementing the auto-segmentation algorithm itself on the server-side, resources can be allocated as necessary to ensure robustness. Scalability is also achieved by decoupling the server from the client, as it can be easily replicated across multiple servers or sites. Images are retrieved and sent using the DICOM standard. To provide more control and customization ability HTTP requests are also used for communication between the client-side and the server-side.

Results So far, a cardiac and a bronchial tree auto-segmentation algorithm have been implemented clinically at Liverpool and Macarthur Cancer Therapy Centre using this framework. An extension was made for MIM (MIM Software Inc., USA) to act as the client side and a similar extension is currently being developed for RayStation (RaySearch Laboratories AB, Sweden). Conclusion The framework presented can help overcome the challenges of implementing an auto-segmentation algorithm in the clinic. It has proven to be useful beyond deploying auto-segmentation tools in the clinic, such as in data mining research projects.

Introduction This work presents a methodology to efficiently associate CT slices with a standardised coordinate location. A reference geometry is defined along the patient superior-inferior axis using the typical distance between several anatomical landmark locations. An image recognition neural network was trained to associate slice appearance with a value corresponding to these reference locations. In this manner, images of an anatomical region will obtain similar coordinate values regardless of patient size or DICOM couch location.

Method One hundred CT image series were manually labelled at landmarks including femoral heads (0 mm), kidneys (+246 mm), liver dome (+376 mm), shoulder (+541 mm), and brain (+748 mm). Reference spatial locations were designated based on mean distance between sequential landmarks for all cases. Intermediate slice positions were interpolated linearly to efficiently generate training data for all values between vertex and thighs. Model training utilised a modified version of the Xception neural network [1] to associate image appearance with a scalar value representing reference geometry location. Model loss used the ADAM optimiser to minimise the squared distance between predicted and true slice positions over 100 epochs. Accuracy assessed by k-fold cross validation (5x 80/20 traintest split).

Results The trained location recognition network could reliably associate CT slice appearance with geometry location with an overall accuracy of ±12 mm across all scan regions. Figure 1 illustrates the predicted z-axis location for a typical case plotted against the original DICOM physical coordinate. Additionally, its use in identifying crop regions to assist with organ segmentation is depicted. Conclusion This work shows the potential to use modern image recognition to evaluate CT scan locations and the utility of a standardised coordinate system for medical image analysis tasks. The technique may assist with longitudinal co-registration, organ localisation, and automated assessment of CT scan lengths for purposes of dose optimisation. References/Acknowledgements As applicable. Introduction Clinical trials in radiation therapy rely on high quality contours which meet the trial protocol requirements. Our team is working towards methods to automate clinical trial QA processes to ensure consistency and reduce the amount of time required for manual review. In this study, the authors present an automatic prostate manual delineation assessing scheme ( Figure 1 ) based on transfer learning of a multichannel 3D ResNet [1] that identifies the inaccurate contours which require major correction.

Method Retrospective data from the prostate SBRT trial PRO-METHEUS (ACTRN12615000223538) was utilised for this study. These data were T2-weighted magnetic resonance (MR) images including five atlas (training) cases containing gold standard prostate contours drawn by 5 expert oncologists and seventy-six trial (testing) cases containing manual contours marked as either correct or incorrect by expert observers. A multichannel 3D ResNet-18 was used to classify the manual contour as accurate or inaccurate and requiring major correction. The multichannel input consists of the MR image in the first channel and the manual contour in the second channel. Due to the limited number of training cases, the ResNet was firstly trained on another prostate MR dataset with 313 cases and then transfer learning was carried out on the five atlas training cases in our clinical radiotherapy dataset. There were no failed contours in the five goldstandard training cases, therefore, inaccurate contour samples were generated by shifting, scaling and flipping the accurate contours during training.

Results The proposed assessing system achieves a classification accuracy of 0.74 with a sensitivity for identifying incorrect contours of 0.73 and a specificity of 0.75. Conclusion The proposed multichannel 3D ResNet model can automatically review manual contours and identify those requiring major correction with a satisfactory sensitivity, introducing a promising approach for improving the efficiency and consistency of clinical trial quality assurance. Introduction Manual expert review of contours for all patients enrolled in radiation therapy clinical trials prior to treatment is logistically arduous and expensive. This has led to the investigation of automated methods for improving contouring accuracy. In this study, the authors present a prostate segmentation method using multichannel AtlasNet [1] ( Figure 1 ) which aims at generating trial atlas guided segmentations and improving manual contours which do not meet the clinical trial protocol.

Method Retrospective data from the prostate SBRT trial PRO-METHEUS (ACTRN12615000223538) was utilised for this study. These data consisted of 81 T2-weighted magnetic resonance (MR) images. There were five atlas (training) cases containing gold standard prostate contours drawn by five expert radiation oncologists, and 76 trial (testing) cases containing manual contours labelled as passing or failing quality assurance. An AtlasNet architecture with multichannel input was trained to segment the prostate. The multichannel input contains the MR image in the first channel and the manual contour in the second channel. The AtlasNet consists of 4 stages: registration (rigid) to each atlas space to reduce anatomical variation; unet segmentation; inverting the rigid transform to map the segmentation back to the original image space; and segmentation fusion. Due to the limited size of the atlas training dataset, the multichannel AtlasNet was initially trained on a larger dataset with 313 cases and transfer learning was then performed on the atlas training cases. Results The proposed system achieves an average Dice similarity coefficient (DSC) of on the whole trial set. For contours (16 contours) labelled as failing by the expert oncologists, the system improves the mean DSC between the failing and corresponding passing contour from to . Conclusion The proposed system demonstrates accurate automatic segmentations guided by gold standard clinical trial atlases and improves failed contours of prostate.

Introduction Radiotherapy is commonly indicated in prostate cancer treatment. Biologically based treatment planning may improve disease-free survival. This relies on accurate biological specification of the tumour. Our goal is to build a statistical biological atlas, to be used with in vivo multiparametric MRI to aid biologically based treatment planning. Method A cohort of 70 men receiving a radical prostatectomy had in vivo and ex vivo MRI of the prostate. Tumour grade and location were marked on whole-mount histology sections. Deformable image registration (DIR) was used to combine these data in ex vivo MRI canonical space and generate a 3D distribution of tumour location and characteristics.

Results The atlas highlighted the persistence of disease in peripheralposterior prostate zones, with voxel-level probabilities for disease presence reaching maxima of 67% ( Figure 1 ). Conclusion Statistical atlas construction with ex vivo MRI overcomes challenges with co-registration of histology sections. A registration Phys Eng Sci Med framework with in vivo MRI can be used to apply this atlas to future patients for accurate biologically-optimised dose escalation.

Introduction Patient-specific quality assurance (PSQA) requires considerable resources while its ability to detect clinically significant errors has been debated [1] . This work investigates the sensitivity of clinically used PSQA procedures to detect purposely introduced, small but clinically relevant, treatment delivery errors. 

Introduction The aim of this work was to investigate an EPID-based patient-specific quality assurance (PSQA) procedure to determine the influence of planning parameters on the results. Method The method uses cine images acquired in-air during VMAT to estimate 3D delivered dose in a virtual cylindrical phantom (VCP) to compare to the treatment planning system dose. Images were acquired using dedicated PCs with framegrabber cards and in-house software. Five linear accelerators (4 Clinacs, 1 Truebeam) and *900 patient plans over a 6 year were analysed using batch calculations with multiple gamma criteria with a 10% of max dose threshold for inclusion. For each plan, 37 separate plan complexity metrics (PCMs) were calculated [1, 2] . These PCMs, alongside categorical linac and planning data were used as inputs in a neural network to predict gamma-pass-rates (GPR) and mean gamma values. The importance of each PCM as well as TPS Version, linac energy, fluence mode and type were obtained from the network and the relative importance of the categorical linac and planning variables were found to be in good agreement with inferential statistical tests performed.

Results Datasets were separated into Clinac (6X) and Truebeam networks due to the different EPID/MLC, energy and fluence modes and plan types used. The models predict GPRs to within *2% at 3%, 2 mm criteria and show high correlation with mean gamma values. The variable importance ranks were different for Clinac and Truebeam with MLC based PCMs dominating, however for Clinac model the aperture area was the most important variable, and for Truebeam the aperture area weighted by MU. For Truebeam model beam energy was also an important factor. Conclusion Aperture area was found to be the most important factor influence results of the PSQA procedure. Further investigations are required to determine if this is planning system, linac or PSQA method related (Fig. 1) . References O029 Investigation of an improved method for EPIDbased IMRT and VMAT patient-specific quality assurance Introduction The aim of this work was to develop an improved model for EPID-based pre-treatment patient-specific quality assurance (PSQA) and investigate the effect of model factors.

Method The VIPER (Virtual Phantom Epid Reconstruction) method uses images acquired in-air combined with kernels to estimate fluence and then dose in water to derive 3D combined field delivered dose distributions in virtual water-equivalent phantoms [1] . A new depthdependent dose calculation algorithm was developed that can derive dose in flat, cylindrical or spherical phantoms. Images were acquired using dedicated PCs with framegrabbers and in-house software for 5 linear accelerators (4 Varian Clinacs, 1 Truebeam) and 905 patient plans over a 6 year period. These were analysed using batch calculations with multiple 3D gamma criteria and 10% of max dose threshold for inclusion. The influence of method-specific parameters was investigated including EPID arm backscatter correction, EPID sag with gantry angle correction, angle subtended by each cine image, low-dose threshold and virtual phantom size.

Results The backscatter correction applied to Clinac images resulted in a 1.3% and 4.8% increase in mean GPR at 3%, 2 mm and 2%, 1 mm criteria respectively. The EPID sag correction was 0.6% and 6.2%. Doubling the number of frames for each image (3.2 to 6.4 degrees average angle) gave 0.2% and 1.5% decrease. GPRs decreased with increasing low dose threshold however sensitivity to PTV dose changes improved. Increasing VCP diameter from 20 cm to 30 cm improved some very large field plan results (Fig. 1) .

The new VIPER algorithm results in very high GPR for standard criteria allowing the use of more sensitive criteria. Analysis of large numbers of patient plans enables the effect of method specific parameters in the PSQA process to be understood and optimised. [1, 2] . An evaluation and test of the recently available Integral Quality Monitor (IQM) transmission detector is presented in this work, focussing on its ability to detect errors in photon beam delivery over a range of clinically relevant conditions. Method The dependence of the measured signal on symmetric and asymmetric square field sizes from 1 9 1 cm 2 to 30 9 30 cm 2 was investigated. The IQM's ability to detect errors in MLC defined field size and field position was characterized. Additionally, its ability to detect delivery errors introduced to clinical VMAT plans for two different sites was assessed. This included measuring 103 nasopharynx VMAT plans and 78 lung SBRT VMAT plans with introduced variations in gantry angle, collimator angle and MLC field size and shift. Lastly, the IQM sensitivity was compared to that of the Arc-Check detector.

Results Differences between IQM calculated (expected) and measured signals remained within ±2% for all symmetric field sizes used, while the asymmetric fields had significantly greater signal variation, particularly for the smallest asymmetric field sizes. The IQM showed higher sensitivity to introduced field size errors than to field shift errors for all the delivered fields and clinical VMAT plans investigated. The IQM was not able to detect collimator angle or MLC shift errors, or small gantry angle errors for the two sets of clinical plans considered. The IQM sensitivity was comparable to the ArcCheck for lung SBRT, but worse for the nasopharynx VMAT plans.

Conclusion The unique design of the IQM introduces dose response characteristics that differ from conventional dosimetry systems and care must be taken when using it for plan verification or online monitoring, especially for small fields that are off-axis in the detector gradient direction.

Introduction Modern radiotherapy plans rely on MLC to define fields and create modulation. This work contrasts two popular TPS MLC models to determine i) the model uncertainties and ii) their impact on different classes of plans.

Method A base linac model was generated in Eclipse 13.6 and Pinnacle 9.6 for a 'golden data' Varian Clinac with Millennium MLC. A range of models were then created in both TPSs, varying leaf transmission (0, 1 and 2%) and modelling of the round leaf ends (Eclipse DLG = 0, 1 and 2 mm; Pinnacle round end offset table intercept shift by 0, 1 and 2 mm). Varied focal source models (0, 0.5 and 1 mm) were also investigated. Each beam model was used to calculate a series of fields, including small static and picket fence IMRT fields, and conformal and high-modulation VMAT arcs.

Results Varying focal source size gave small changes in profiles, with, Pinnacle being less sensitive than Eclipse. Larger differences were observed for leaf transmission, with varying impact across plan classes.

Here, Pinnacle was more sensitive than Eclipse, especially for the highmodulation VMAT where a near-uniform dose difference was observed in-and out-of-field. Round end modelling had the largest impact on plans, with the picket fence profiles demonstrating different behaviours for abutting segments between systems (picket junction differences of +40% and +25% in Eclipse and Pinnacle respectively between models). High-modulation VMAT field dose changed in unpredictable ways, with local differences ranging from -7% to +8% and -12% to +7% in high and low dose regions respectively between models. Conclusion Using this bank of fields and beam models, the significant differences in MLC modelling between systems and between plan classes is described. The modelling of MLC in the treatment planning system is of critical importance to generate accurate and deliverable plans.

O033 What is wrong with Eclipse's Dosimetric Leaf Gap parameter?

Introduction Eclipse TPS uses the Dosimetric Leaf Gap (DLG) parameter for modeling the extra transmission through MLC's leaves' tips. Its experimental determination is simple and precise [1, 2] . The TPS beam model, including DLG values, can be used to predict doses happening on experimental determination of DLG, this provides a consistency test. We found the DLG calculated after TPS predicted doses perfectly match the fed values (see Table 1 ). Nevertheless, several publications [3] [4] [5] [6] [7] point to the need of increasing DLG values to correctly predict doses in cases of highly modulated deliveries. A suboptimal tongue-and-grove modelling has been pointed as main cause of discrepancy [3, 7] . This study further investigates the possible reasons for this contradiction. Results Experimental DLG values agree with predicted doses better than 0.5%. Agreement between Eclipse predictions and measurements when using static MLC and low-modulation VMAT were found to be better than 1%. However, highly-modulated deliveries showed output discrepancies as high as 5%, intriguingly, displaying almost identical relative dose distributions (see Figure 1 ). These discrepancies got reduced to less than 1.5% when DLG was increased by 0.5 mm; with no significant on the shape of the dose distributions. Figure 1 Eclipse predicted dose distributions, corrsponding to a brain SRS treatment type (Audit), Calculated using two different DLG values: Red distributions correspond to measured DLG, Blue distributions correspond to increased (+0.5mm) DLG. Blue distributions on right side panels have been renormalized by -2.9%

Conclusion We think the need to increase DLG is rooted on two concurrent problems: The published deficiency on tongue-and-grove modelling and a deficient AAA/Acuros source model. Results Separating OAR local dose variations across the ACDS national audit data by algorithm, in Figure 1 , suggests variability in accuracy between algorithms at predicting OAR dosimetry. Sub-optimal penumbra modelling is a potential source of OAR dosimetric discrepancy, and can be varied by increasing ECUT. ECUT effects to an OAR with a wrap-around target were modelled in DOSXYZnrc/BEAMnrc. The transverse plane comparisons are presented below in Figure 2 .

Conclusion ACDS audit data suggests evidence for variability in the accuracy of different algorithms to predict OAR dosimetry, with the ECUT effect on penumbra modelling being identified as one potential source of this discrepancy.

Stereotactic and other small-field treatments will have larger penumbra-to-field ratios, and thus the penumbra modelling effects on OARs may be significantly more pronounced. Clinical implications may be minimal for large-field treatments, such as for prostates. References/Acknowledgements Introduction Titanium implants cause significant dosimetric perturbations in radiotherapy. Excessive dose at the tissue-implant interface compromises osseointegration, whilst dose shielding may compromise treatment. An alternative material is the PAEK polymer family, but it requires detailed dosimetry characterisation at the interface.

Here we use Monte Carlo (MC) methods to characterise the dosimetry at the tissue-implant interface of titanium and PEEK in comparison with bone.

Method The Monte Carlo simulation is designed to score the absorbed dose in 2 mm slices for the whole phantom and 0. 2 mm slices near the implant to generate a depth-dose curve ( Figure 1 ). Radiochromic film was used to measure the dose at the water-implant interface. The dose was also calculated with the Varian Eclipse TPSv15.6.

Results The Monte Carlo calculated dose perturbation is shown in Figure 2 . At the incident interface, a 34% dose enhancement is observed for titanium, 12% for bone and 6% for PEEK. At the exit interface, an 18% dose decrease is observed for titanium, but a dose enhancement of 7% for PEEK and 3% for bone. There is a dose reduction of 15% for titanium and 3% for bone 5 cm beyond the exit interface, but enhancement of 0.3% for PEEK.

Conclusion Our results show that in a radiation beam, PEEK behaves more like bone than titanium does. By avoiding an over-response at the entrance interface (*30%), osteo-integration is not compromised, while the absence of a cold spot (*18%) at the exit interface avoids treatment failure. Film dosimetry confirms interface dose perturbation, while treatment planning underrepresents the potential impact to patients. O037 Reducing sensitivity to asymmetric scattering conditions in source-tracking based treatment verification of HDR brachytherapy Introduction Our treatment verification system [1] utilises a flat panel detector (FPD) mounted in the treatment couch to track source positions by capturing source radiation exiting the patient. Previously reported in vivo tracking capabilities have exhibited measurement uncertainties due to asymmetric scattering in patient geometries [2] .

Here, we show that asymmetric scattering affects the radiation distribution at the FPD and hence the source position measurement. We demonstrate two methods to compensate for the asymmetric scattering conditions. Method Source tracking in a phantom was conducted in symmetric and asymmetric scattering conditions, by varying proximity to air interfaces, compared to ground-truth derived from subtraction radiography [1] . Inaccuracies caused by asymmetry were mitigated using two techniques: physical by introducing an anti-scatter grid to improve the primary-scatter-ratio; and computational by adopting a data-driven optimisation of our algorithms that extract the source position from the FPD dose distribution.

Results Source position accuracy decreases from \1 mm under symmetric conditions to an error of up to 9 mm in the most asymmetric case, thus confirming the influence of asymmetric scattering.

Optimising tuneable parameters of the position determining functions reduced their sensitivity to asymmetry, with differences improving from\9 mm to\2. 5 mm. Employing the anti-scatter grid to improve the primary-scatter ratio, discrepancies improved from \9 mm to \4 mm (see Fig. 1 Introduction Avoiding the risks and costs involved with the surgical insertion of radio-opaque fiducials, several commercial and academic markerless lung target tracking approaches have been developed to further improve patient safety during lung cancer stereotactic ablative body radiotherapy (SABR). However, these approaches had yet to be benchmarked using a common measurement methodology. This knowledge gap motivated the Markerless Lung Target Tracking Challenge (MATCH) [1] . Method The MATCH is an American-Association-of-Physicists-in-Medicine-(AAPM)-sponsored Grand-Challenge in 2019/2020. The Figure 1 Difference between measured and true positions relative to the inferior phantom-air interface. Two grid results are shown against no-grid data emphasizing that grid choice is important. Error bars represent Min-Max range over 6 measurements Figure 1 Overview of the AAPM MATCH Grand Challenge Phys Eng Sci Med participants aim to accurately and precisely localise lung targets with time in (Part-A) a retrospective in-silico study and (Part-B) prospective phantom experiments. Common to both parts are a 3Dprinted lung phantom including three lung targets [2] , and a lung SABR planning protocol [3] . The phantom is moved rigidly with patient-measured lung target motion-traces, which also act as groundtruth. In Part-A a volumetric-modulated-arc-therapy-treatment is delivered to the phantom programmed with four unknown motiontraces. A dataset consisting of treatment-planning data and intratreatment kV-and MV-images is provided to the participants. In Part-B the participants use their own approach and workflow to localise the target during the dose delivery for five motion-traces. All participant submissions will be analysed and ranked based on the percentage of the tracking error values within 2 mm of the ground-truth. The challenge is open to any participant, and participants could complete either one or both parts (Fig. 1) .

Results More than 20 institutions registered for MATCH. At the time of the abstract submission, eight results from five institutions in three countries have been received. The results of the challenge will be made available after the deadline 31st of August 2020 and presented at the EPSM conference 2020.

Conclusion A common methodology for measuring the accuracy of markerless lung target tracking algorithms has been developed and used to benchmark academic and commercial approaches retrospectively and prospectively. Introduction Fiducial marker seeds are often used as a surrogate to identify and track the positioning of prostate volume in the treatment of Prostate cancer. This research is undertaken to develop and validate a real time deep learning object detector using a YOLO v2 Convolution Neural Network (CNN) that can detect and track fiducial marker seeds in prostate cancer patients. This detector does not require any prior knowledge of the seed positions. A software program has been developed that can visualize each image obtained directly from the kV XVI panel, draw a bounding box around the seeds and plot the centroids of the seeds in real-time.

Methods The detector was trained using a variety of datasets of fiducial marker seeds, that were initially labelled and used as ground truth. The validation and testing were performed on unseen XVI images. Software program was developed to display the projection images and predict the seeds using YOLO v2 and determine the centroids on each image while scanning the directory containing image files.

Results The fiducial marker seeds were successfully detected in 98% of images from all gantry angles, the variation in the position of the seed centroid was within ± 1 mm. The percentage difference between the ground truth and the detected seeds was within the 3 % tolerance.

Conclusion The deep learning model to detect fiducial marker seeds in kV XVI panel images with no prior knowledge of their positions was successfully demonstrated. This is an ongoing project at the Princess Alexandra hospital and work is underway to extend it to other sites for tracking moving structures with minimal effort. The final aim is to incorporate it into clinical practice for use in radiotherapy treatment.

O040 Surface image guided radiation therapy implementation Introduction The use of optical methods to assess the patient surface is the basis of a range of commercial solutions for patient setup and monitoring in Radiation Therapy simulation and treatment. Following a market analysis by a multidisciplinary team our department has invested in a C-Rad Sentinel system and four C-Rad Catalyst+HD systems.

Method The commissioning is based on AAPM TG147 [1] with input from recent work [2] . This included the assessment of the integration of peripheral equipment, measurements of spatial reproducibility, evaluation of static and dynamic localization accuracy and the development of ongoing quality assurance procedures. Commissioning was tailored to the intended use of the systems, which includes DIBH and 4DCT for the Sentinel on a CT scanner and DIBH and SRS for the Catalyst+HD on linacs, in addition to patient setup and monitoring for a broader patient population. A Hexamotion phantom with a custom platform [3] was used to assess static and dynamic motion responses. For the field of view (FOV) evaluation a novel mapping method has been developed. Results At the time of submission the Sentinel and one Catalyst+HD system had been installed, the commissioning measurements for the Sentinel were complete and commissioning of the Catalyst+HD had begun. Sentinel warm up from a cold state showed isocentre drifts of 1.4 mm in the up-down direction and 2. 3 mm in the sup-inf direction with stabilization at approximately 100 minutes. Warmup from the recommended standby was within 0.12 mm.

The measured height of a stationary object on a moving CT couch fluctuated by about 1 mm. Reconstruction of 4DCT images was flawless (Figs. 1, 2).

Conclusion Sentinel is ready for clinical use. As recommended, it should remain in standby overnight, with warmup needed in case of power outage. Introduction Motion of the tumour during a radiotherapy treatment can compromise the clinical outcome for the cancer patient. To address this challenge, the suitability and performance of three different models of the Intel RealSense TM depth camera technology have been investigated as potential tools for measuring the respiratory-like motion of a surface.

Method Three different models of Intel Realsense TM depth camera (the D415, D435, and SR305) were used in this study (shown in Fig. 1a ). Acquisitions were performed with in-house software developed using the MATLAB wrapper functionality of the Realsense TM SDK 2.0 at a frame rate of 30 fps [1] . The precision of the camera depth data has been characterised as a function of the measurement distance up to 1.2 metres from a stationary and moving flat surface. Additional measurements, with the three cameras were performed for a moving thorax phantom to simulate a more clinical scenario, shown in Fig. 1b . Results Figure 2 (a) shows the standard deviation over 300 frames of the depth measurement as a function of measurement distance (depth) for a stationary surface. Figure 2b -d shows examples, for the three cameras, of the measured displacements of a surface moving with a sinusoidal motion amplitude of 2.5 mm and frequency of 0.25 Hz. Similar results for the three cameras were observed when measuring the motion of the thorax phantom surface. Conclusion The D415 and SR305 cameras were able to measure the distance between the camera and surface with a similar precision that deteriorated as the distance increased. The D435 camera showed significantly worse precision with far higher levels of noise evident in the measured depth values. The D415 and SR305 cameras could be The reasons for modifications include broadening the sites, optimising dose and improving the image quality. Although there has been substantial development in radiology [2, 3] to optimise of exposures, the development of imaging protocols for paediatrics radiation therapy are not widely published. A review of ROPART exposure settings were compared with recently published low-dose protocols by Ryan et al. [4] . Method Low-dose protocols for paediatric IGRT was determined through a phantom study. Three anatomical sites, head and neck, thorax, and extremity were investigated. For each site the kV and mAs were adjusted until the image quality was adequate to perform image matching, as judged subjectively by a group of experienced radiation therapists. In addition, the relative dose comparison by site between ROPART and Ryan et al [4] protocols were assessed using RaySafe Unfors X2 with RF sensor at isocentre with a 20 9 20 cm 2 field. Purpose Stereotactic ablative body radiotherapy (SABR) has recently emerged as a promising approach for treatment of refractory ventricular tachycardia [1] . We report Australia's first application of SABR for refractory ventricular tachycardia.

A 75 year old man with severe, dilated cardiomyopathy with recurrent VT who had underwent three catheter ablation procedures was referred for SABR. Targeting was performed using a combination of invasive mapping from catheter ablation procedures, 12 lead ECGs, and using novel 3D non-invasive 252 electrocardiographic imaging (CardioInsight, Medtronic) process. A 4DCT radiotherapy simulation CT was performed with IV contrast. Target definition was performed by cardiologists and radiation oncologists in collaboration with international experts at University of Washington in St. Louis. A 3 mm planning target volume (PTV) expansion was applied to the target volume. A single fraction of 25 Gy was planned to cover 95% of the PTV, with sharp reduction of dose outside the PTV.

Results The target volume (right ventricle wall) was 17 cm 3 , with the treated PTV 47 cm 3 . The treatment was delivered with the presence of cardiologists, radiation oncologists with monitoring of the patient by cardiac electrophysiologists and anaesthetics team. Image guidance was performed using cone beam CT, using cardiac artery calcifications to assist with image registration. The treatment delivery time was 25 mins, which included two CBCTs prior to treatment and one mid-treatment, and 7 mins beam on time. There were no acute post-treatment toxicities observed.

Conclusion Single fraction SABR was successfully delivered to the right ventricle wall for treatment of refractory ventricular tachycardia. A strong inter-disciplinary and international collaboration was fundamental to the successful completion of the procedure. Further work will focus on spatial combination of treatment planning CT geometry with imaging and electrical mapping used for targeting. References Introduction GFR estimation using radiopharmaceuticals is prone to sample collection, counting and calculation errors, which are irreversible post-calculation, but easily rectifiable if identified in realtime. Results vary between radiopharmaceuticals, calculation methods and correction formulas. A new spreadsheet with real-time feedback based on quality control checks and a report with relevant age-specific normal ranges has been developed and implemented.

Method QC checks were based on departmental protocol for activities, timing and time sequence, and expected counts from standard activity. Tolerances were based on equipment precision with allowances for human error. Normal GFR ranges applicable to this calculation method were obtained by recalibrating published reference ranges and the compatibility with our patient cohort was assessed. 957 past reports were analysed and common errors were identified. The reported normal range was found to be significantly different from widely accepted publications. A presentation to staff was followed by active assistance in the initial months of implementation.

Results The most common errors such as typos, time-sequence errors and missing digits in counts have been completely eliminated.

Technologists immediately rectify flagged data-entry fields before proceeding further. Any breach of sample counting tolerances have not yet been encountered. Taking all the tolerances into account, \5.8% error in reproducibility has been estimated. Confidence in counting performance and quantification of reproducibility allow a significant dose reduction of 50% in adults and up to 87% in children.

Conclusion The quoted normal range in our report, adapted to the method used, ensures reporting uniformity and increases referrer reliability to make informed clinical decision. Further data accumulation will allow stringent tolerances and evaluation of postcalculation QC metrics like volume of distribution and clearance halftime present in the spreadsheet. One challenge with transitioning to a radically new protocol in a large multi-site institution has been to get all staff to adapt to the changes simultaneously. Acknowledgements I would like to acknowledge Thomas Greig for his help with data consolidation and Nicholas Hille for his support to implement this change in the department. Introduction Internal dosimetry is essential for determining thresholds for decreased organ function from therapies and quantifying radiation dose detriment to the public from diagnostic imaging. The process of calculating doses from functional imaging is time consuming, requiring accurate image segmentation and data analysis. The following work presents the development and implementation of an automated internal dosimetry application utilising convolutional neural network (CNN) based image segmentation of low dose CT scans.

Method Two hundred unique whole body low dose CT images were manually segmented to create a training dataset containing binary masks of the lungs, liver, spleen, kidneys, brain, heart and vertebrae as ground truths. CNNs were constructed in Tensorflow1.14.0, trained with an augmented instance of the training dataset and utilised in a segmentation algorithm. The segmentation algorithm was incorporated into a dicom server framework running on an external network location which receives CT data as a PACS location and returns the segmentation in the form of a radiotherapy planning structure. An additional twenty whole body CT scans were segmented manually to evaluate the segmentation algorithm. Results All segmentations for the twenty evaluation scans produced volumes within 90% of the manually segmented volume. The mean processing time of the twenty evaluation segmentations was (mean SD) seconds. Figure 1 slices of a whole-body CT scan segmented with the automated algorithm. Utilising a radiotherapy planning structure output allows the end user to dynamically edit the segmentation allowing for manual optimisation if required.

Conclusion An automated CNN based segmentation algorithm has been developed in a framework which is easily implemented clinically. The algorithm presented demonstrates a significantly more efficient alternative for dosimetric analysis. Further evaluation and packaging into a user-friendly application will be done in the immediate future.

O047 Automatic segmentation of metastatic prostate lesions in PET/CT images using convolutional neural networks Introduction The application of radiomics-based predictive models of advanced prostate cancer using 68 Ga-PSMA PET/CT images requires segmentation of metastatic prostate lesions. Manual segmentation is a time consuming and laborious process when there are tens or hundreds of lesions. The automation of this process could reduce user time required for this process, facilitate efficient treatment planning for patients undergoing radionuclide therapy as well as minimise inter-operator variability. This study aims to develop an automated segmentation framework using deep learning techniques for metastatic prostate lesions in whole body 68 Ga-PSMA PET/CT images. Method A local dataset of 256 patients with metastatic prostate cancer, who have had whole body 68 Ga-PSMA PET/CT assessments, was analysed. A clinical expert manually delineated PSMA-avid lesions in each scan, which constituted the ground truth segmentation data that was subsequently used to train a convolutional neural network (CNN). The metric of model evaluation was the Dice similarity coefficient (DSC), measuring the overlap between the model predictions and the ground truth manual delineations. Results Preliminary results suggest the feasibility of whole-body PET/CT automated delineation using deep learning techniques. Utilising the MultiResUNet CNN model architecture [1] , the best average DSC achieved thus far is 60.5%, which is comparable to literature results [2] . The model will continue to be refined with a combination of pre-processing steps and optimised CNN architecture (Fig. 1) . Conclusion This novel research demonstrates the feasibility of automatically segmenting metastatic prostate lesions in whole body PET/CT images using deep learning techniques. This could facilitate efficient lesion detection and quantification, allowing medical imaging specialists to focus on interpretation and comparison tasks and optimising workflow, and can alleviate some of the laborious manual work of lesion detection. Introduction This study is being undertaken to utilise spatial dynamic characterisation of uptake of 18 Fluoro-O-(2) fluoroethyl-Ltyrosine ( 18 F-FET) to generate biomarkers predictive [1] of glioblastoma (GBM) treatment outcome.

Method A Patlak graphical analysis technique was developed and tested on 24 18 F-FET positron emission tomography (PET) patient image data sets. The dynamic images were acquired from approximately five minutes post tracer injection, with images taken every minute for between 20-30 minutes in total. Allowing for the uptake flux (Ki) of the radiotracer into the brain tissue and time-to-peak (TTP) of tissue time activity curves to be calculated on a voxel-wise basis, forming 3D parametric maps which can further characterise cancerous regions of the brain. 18 F-FET PET scans were performed on GBM patients with both post-resection/pre-radiochemotherapy and post-radiochemotherapy scans. The superior sagittal sinus vein of each brain scan was segmented using the earliest PET frames of high plasma-blood activity using 3D Slicer in order to generate an input function for each patient. Tissue time activity curves were extracted on a voxel-by-voxel basis in MATLAB, allowing a graphical analysis to be implemented and uptake flux and TTP maps of the brain to be calculated. The intra-patient stability of these maps was tested in 8 patients who had a third series of pre-radiochemotherapy scans taken a week apart by measuring mean square error statistics between the maps.

Results Maps of parameter uncertainty were generated by extracting confidence intervals and mapping their spatial variations. Voxel-level time-activity relationships were found to be strongly impacted by large amounts of image noise, greatly restricting the ability to derive both dynamics parameters as well as time-activity curve features (time-to-peak and categorised patterns).

Conclusion The presented approach for calculating parametric maps of dynamic 18 F-FET PET scans allowed for enhanced spatial characterisation of glioblastoma by highlighting areas of increased uptake flux and early TTP, with dynamic image noise being a major source of uncertainty. 

Introduction Stereotactic ablative body radiotherapy (SABR) is a novel treatment for patients with inoperable renal cell cancer. We have previously demonstrated that delivered dose to kidney during SABR is associated with renal function decline [1] . Typically these patients have impaired kidney function, and occasionally a solitary kidney, therefore reductions of ipsilateral kidney dose is desirable. This study determines the reduction of ipsilateral kidney dose from elimination of respiratory motion in kidney SABR.

Method For twenty-nine kidney SABR patients, the gross tumour volume (GTV) was segmented on each phase of a phase-binned 4DCT. Tumour motion was obtained from the GTV centroid position on each phase. Motion managed (MM) plans were optimized and calculated to the GTV on the exhale phase. Non-motion managed (NMM) plans were optimised and calculated on the average of the 4DCT to an internal target volume (ITV). The NMM plan was copied ten times to the exhale phase, with isocentre corresponding to the tumour motion in that phase. The dose was calculated in each phase and averaged. Change in ipsilateral kidney volume receiving 50% of the prescription dose (V50%) was assessed as a function of tumour motion amplitude (TMA). Pearson correlation coefficient was used in the statistical analysis.

Results The mean±st.dev TMA was 6.8±3.4, 3.2±1.7, 1.4±1.1 mm in sup-inf, ant-post and left-right directions respectively. A linear relationship between GTV to ITV volume difference and TMA was observed (R 2 = -0.89; P-value.001) as shown in Figure 1 . A larger volume of non-tumour ipsilateral kidney was spared in MM plan compared with NMM plan (V50% reduction of 11.0%/4.6%/0.7% Phys Eng Sci Med max/median/min). V50% decrease is correlated with TMA as shown in Figure 2 (R 2 = -0.67; P-value \ 0.001).

Conclusion Removing respiratory motion in kidney SABR reduces dose to the ipsilateral kidney, with the benefit increasing with increasing target motion. Results We developed a clinical framework to address different components of SBRT process which include: patient assessment, simulation and immobilization, motion management, target and normal organ delineation, treatment planning, quality assurance, participation in external audit, treatment verification imaging and delivery, and patient follow-up. The 4D CT checklist and pre-treatment checklist for SBRT lung are shown in figure 1 and 2, respectively.

Conclusion SBRT is a resource intensive treatment technique and is specific to individual treatment centre. The effective and satisfactory transition from conventional radiotherapy practice to hypo-fractionated SBRT requires additional attention. Implementation of this cutting edge technology at CWCCC has made hypo-fractionated O052 An attempt to quantify the effect on output factor of a sub-millimetre variation in the size of a small radiation field Introduction In megavoltage photon beams, dosimetry of radiation fields B 30 mm presents several challenges [1] . Besides the uncertainty on positioning the dosimeter [2] , and on correction factors required to relate its reading to dose [1] , uncertainty in the effective Method We eliminated the influence of other compounding uncertainties (on positioning and on correction factor) by using a 2D solidstate array, the Duo, which has a 0.2-mm resolution over 5.2 9 5.2 cm 2 , and does not require corrections [3, 4] . We placed the Duo at 90-cm source-to-surface distance, 10-cm depth in Solid Water, and aligned it with the central axis of an Elekta Versa HD using a 5-mm square field [5] . We then delivered 6 MV FFF beams, collimated with an Agility multi-leaf collimator in the cross-plane and with the Y-jaw in the in-plane, applying sub-millimetre shifts to the collimators and measuring OPF each time. The equivalent field was defined as EF=H(A9B), where A and B were the cross-plane and in-plane FWHM.

Results OPF are in Table 1 , normalized to the 10 9 10 cm 2 reference field. In a 5.7-mm EF, a cross-plane change of 0.2 mm resulted in a 5.0% variation in OPF, whereas in a 20.0-mm EF the same change resulted in a 0.5% variation. In fields equal to the same EF, collimated with interchanged configuration of MLC/Y-jaw, OPF varied by 0.7% in the 5.6-mm EF, and by 0.1% in the 10.3-mm EF.

Conclusion OPF in small stereotactic beams should be reported as a function of effective field size [5] . Here, we proposed that the effective field size should be reported with sub-millimetre accuracy, in both cross-plane and in-plane directions.

Introduction Deep learning methods have proven their value in radiology, from segmentation to equipment maintenance. Recently, the prospect of utilising deep learning to aid image reconstruction has been commercially realised. Manufacturer claims include features such as improved noise texture and better performance at low doses. However, deep learning image reconstruction (DLR) is fundamentally different from other reconstruction technology and introduces the potential for novel artefacts [1] , so appropriate care should be taken in assessing and commissioning it.

Method The properties of the DLR algorithm in comparison to iterative reconstruction (IR) and filtered back projection (FBP) were assessed using a CatPhan 600. Noise images were evaluated by subtracting images reconstructed using each algorithm from an averaged image acquired by performing repeated FBP, which is taken as the ground truth. These images are used to evaluate the noise power spectrum (NPS), as well as illustrate differences between the algorithms, demonstrating any artefacts introduced by DLR. The modulation transfer function was also measured to assess spatial resolution.

Results The DLR algorithm demonstrated improved noise frequency and similar spatial resolution. The relationship between noise and dose was similar for IR and FBP, however the response of DLR at very low doses appeared less consistent than for the established algorithms. Inspection of the subtracted images showed that the DLR differed from the FBP standard by more than IR on average (mean HU difference of 0.07 for DLR vs 0.01 for IR). Furthermore, the outline of structures of the CatPhan were visible in the subtracted images (Figure 1 ), indicating that there are circumstances in which spatial fidelity is subtly affected.

Conclusion The value of the DLR algorithm was demonstrated, and echoes results found by other investigators [2] . However, the importance of comprehensive quality assurance when using novel algorithms was also clear. Introduction Image Guided Radiotherapy (IGRT) provides the ability to accurately target the radiation to the tumour volume and to minimise the dose to the surrounding normal tissue. Cone Beam Computed Tomography (CBCT) is the most common modality used in IGRT and although the imaging dose is small compared to the therapeutic dose it is delivered to a large volume of normal tissue and should be minimized.

Method The current work investigates the possibility of reducing the imaging dose from CBCT by modulating the mAs during an acquisition to match the patient anatomy. The current hardware in clinical CBCT systems does not allow the modulation of the mAs during a scan and therefore a framework for simulating this via software processing of the projection data was established. CBCT projection datasets were acquired on an Elekta Synergy XVI system at different mAs values. A Python script was then used to combine different proportions of the high and low mAs projections into a new dataset that could then be reconstructed on the XVI system. Results Figure 1 shows a typical slice in the 3D volumetric image of a Brainlab Pelvis phantom after image reconstruction in XVI. The image on the left shows the reconstruction with 100% low mAs projections. The image on the right shows the reconstruction with 100% high mAs projections. The central image was reconstructed with 50% each of the high and low mAs projections. Quantitative analysis of CBCT reconstructions of a Catphan phantom showed a linear trend in the low contrast visibility with the proportion of high mAs projections used in the image reconstruction.

Conclusion The current framework has been shown to be an effective tool for exploring the possibility of reducing the imaging dose from CBCT. It has been demonstrated that it is possible to effectively balance the reduced low contrast visibility with reduced imaging dose by modulating the mAs during the acquisition scan. It has also been demonstrated that new metrics are needed to quantify the artefacts that may be introduced with modulated mAs schemes.

O055 Efficient projection based method for metal artefact reduction in cone beam CT Introduction Cone Beam Computed Tomography (CBCT) is the most common modality used in Image Guided Radiotherapy (IGRT). Metal Artefacts in CBCTs can dramatically reduce the accuracy of image registration in IGRT which can in turn reduce the ability to accurately target the tumour volume and minimise the dose to the surrounding normal tissue. It is therefore vital to develop methods for the accurate and efficient reduction of metal artefacts in CBCT.

Method The currently proposed method extends the projection-based method of Zhang et al [1] . The key improvements in the current methodology is the automatic segmentation of the metal shadows, the use of a highly efficient algorithm for the in-painting of the metal shadow and the addition of a space-carving technique to contour the metal object. The method was tested using a water phantom with high density metal inserts. The CBCTs were acquired on an Elekta Synergy XVI system. Images reconstructed with and without metal artefact reduction were compared with the ground truth images that were acquired with no metal objects.

Results Figure 1 shows a typical slice in the 3D volumetric image of the water phantom after image reconstruction in XVI. The image on the left shows the original reconstruction with severe metal artefacts.

The image on the right shows the corresponding slice in the reconstructed image after metal artefacts reduction. The contours of the metal object derived from space carving have been added. The central image was acquired with no metal objects present and with the metal object contours added was used as the ground truth data for quantitative analysis.

Conclusion The current methodology has been shown to be highly effective in reducing metal artefacts in CBCT. It can be easily incorporated into current clinical workflows with no additional operator interventions or significant increase in time. The addition of the space carving technique allows the use of the metal object contours for image registration. Introduction Cone Beam Computed Tomography (CBCT) scans are prone to negative effects from scattered radiation, resulting in incorrect Hounsfield Units (HU) [1] . Monte Carlo (MC) based scatter correction has been shown to resolve such issues [1, 2] . In such methods, scatter contribution in each projection is determined using MC simulation of projections, allowing acquired CBCT projections to be corrected. In this study we evaluated the impact of various methods of performing scatter correction with MC generated scatter contribution.

Method A pipeline for MC scatter generation and open-source CBCT reconstruction was created using EGSnrc and RTK. A Fan-beam CT (Philips Brilliance) and CBCT (Varian TrueBeam) of an anthropomorphic phantom were acquired. Primary and scatter simulated projections were produced from the fan-beam CT. The scatter was scaled via a ratio of the acquired projection to the MC equivalent (primary + scatter) and removed. Methods of this included scaling the scatter with a per-pixel ratio, the mean ratio of pixels including the phantom in the projection, and the mean ratio of all pixels.

Results Figure 1 contains images of the fan-beam CT, original CBCT and scatter-corrected CBCT from each method. Partial image correction resulted in improved HU consistency at the phantom periphery. This can be seen in the line profiles of Figure 2 . Optimal scatter correction has been achieved via partial ratio scaling, due to improved image uniformity. The HU difference from the centre to the periphery was reduced from 106 HU (original CBCT) to 57 HU with the partial ratio correction. Conclusion A MC based scatter correction pipeline has been implemented. Scaling the scatter with a ratio calculated from part of the projection increased image uniformity and HU. References Introduction Image-guided radiation therapy (IGRT) is an essential component of modern radiotherapy. Cone-beam computed tomography (CBCT) is used to verify patient position and monitor changes in patient anatomy. Protocols provide guidance on the required quality assurance to ensure CBCT systems are capable of providing clinically useful images [1] [2] [3] . It has been reported that protocol-based quality assurance (QA) tolerances are not reliable at predicting clinical image quality issues [4] . CBCT technologies and applications undergo rapid development, meaning that historical QA tolerances may no longer be applicable. Additionally, many protocols instruct the user to compare to baseline data, but do not provide guidance on what variation from baseline is acceptable. Control charts can be used to monitor quality and identify out-of-control results. This work presents a novel approach to determining CBCT QA tolerances that address the above issues using process control charts.

Method We obtained 50 CBCT scans of a CATPHAN 604 phantom over a period of 14 days on a Varian Truebeam using standard Head and Thorax protocols. The data sets were analysed using pylinac [5] . We produced control charts using QATrack+ [6] .

Results A comparison of control limits and protocol and manufacturer tolerances for the Head protocol in Table 1 . using the combined data set of MC and measured values was performed for each chamber, created a set of parameters, a and b, that allows for the calculation of k Q at other beam qualities. Results The consensus data of k Q , combining measured and MC values, for the NE-2571 Farmer-type ionisation chamber is shown in Figure 1 . Consensus k Q data was generated for each of the 23 chambers. The consensus k Q data in both a tabular form as in TRS-398, and as a set of chamber specific parameters a and b, as in equation 1. The combined standard-uncertainty in the new k Q values is 0.62%. Conclusion A new consensus dataset for the beam quality correction factor k Q in high energy photons has been created. It consists of values from many international research groups and standards laboratories. The consensus data will likely form the basis of published k Q values in the updated protocol TRS-398. Australian Clinical Dosimetry Service, ARPANSA, Australia Introduction ARPANSA maintains two free air chamber (FAC) primary standards for low and medium energy x-rays. The low energy FAC (LEFAC) and medium energy FAC (MEFAC) are used over an energy range of 10-100 kVp and 40-320 kVp respectively. All x-ray air kerma calibrations offered by ARPANSA are traceable to these FACs.

Due to the adoption of the ICRU 90 recommendations by the CCRI(I) in 2018, the correction factors of each chamber changes, which in turn affects calibration factors of secondary standards calibrated at ARPANSA. Here, we present this change in calibration factors due to the adoption of ICRU 90. Method Both the LEFAC and MEFAC were modelled in egs_fac [1] and validated against prior Monte Carlo models using monoenergetic source inputs and the same cross sections. [2] Once the Monte Carlo models were validated, the simulations were rerun with renormalised photoelectric effect cross sections, as per ICRU 90. The correction factors were calculated by convolving the monoenergetic data with the spectra of the beams used in calibration. The new correction factor, k ii k w, was also incorporated. Results Both the new LEFAC and MEFAC models showed good agreement to previous models, with differences either within uncertainty or attributable to previous assumptions made for efficiency. The adoption of ICRU 90 data had a small effect on the product of correction factors for both the LEFAC and MEFAC. The LEFAC correction factor had a decrease between 0.2 to 0.4%, depending on beam energy. The MEFAC correction factor showed a change of -0.4% to +0.1%. Conclusion The correction factors for the FAC primary standards at ARPANSA have been recalculated, incorporating adoptions due to ICRU 90. While the small changes that are seen affect calibrations performed at ARPANSA, the differences are within the overall uncertainty budget of the calibration service. Phys Eng Sci Med megavoltage photon beams. With the prospect of proton radiotherapy in Australia, it is time to consider replacing our calorimeter with a version capable of measuring in both photon and proton beams. Some overseas primary standard laboratories have already upgraded, with both graphite and water options used. The relative merits are worth considering.

Method We have explored graphite and water calorimetry through the scientific literature [1] , and experimented at the Australian Synchrotron, where graphite, graphene and water calorimeters have been tested in high-intensity beams.

Results Graphite has the advantage of a low specific heat capacity of 0.7 Gy/°C compared to 4.2 Gy/°C for water, with much greater temperature rise for graphite for the same photon fluence, allowing sensitive measurements at lower-dose-rate 60 Co beams. However, graphite requires conversion from absorbed-dose-to-graphite to absorbed-dose-to-water by Monte Carlo calculation or other method, introducing extra uncertainty. Conversely, water has the advantage of realising absorbed-dose-to-water directly. A water calorimeter is more massive, and therefore less portable. Both require correction factors, and both require embedded thermistors for heating/temperature measurement. These thermistors sit in the beam and are subject to output changes at the onset and after irradiation. This effect is known in calorimeter studies in kV beams (see [1] , Figure 6 ) and is attributed to radiation induced current in the thermistors and leads. Data on this effect will be shown from our calorimeter studies at the Synchrotron. Conclusion Internationally both graphite and water calorimeters are used as primary standards, providing robustness to the international average determination of absorbed-dose-to-water. If ARPANSA chooses water calorimetry for a new primary standard while retaining graphite for comparison, robustness will be added to ARPANSA's capability to accurately measure absorbed-dose-to-water for present and future radiotherapy modalities in Australia.

Introduction Magnetic field correction factors, k B , are required for dosimetry in magnetic fields. Current methods for determining k B include: transfer from water calorimeters or chemical dosimetry, measurements during magnet ramp up/down or monte-carlo simulations. This work aims to validate a methodology for measuring k B via measurements using a conventional linac.

Methods k B is a ratio of the absorbed dose to water calibration factor in the magnetic field to 0 T at the same beam quality (N DwQ B /N DwQ 0T ). k B contains i. dose to water ratio in the magnetic field to 0T (D w B /D w 0T ) and ii. measured charge ratio of a dosimeter at 0T to the magnetic field (M 0T /M B ). This work proposes a cross-calibration between a conventional linac and a MR-Linac (MRL) to determine k B . Charge ratios between a microdiamond and PTW30013 Farmer were measured at 0 and 1T positions on the Australian MRL and on a conventional linac with the effective point of measurement at 10 cm depth in solid water. These measurements enabled: Introduction The Octavius SRS1000 ion chamber array was acquired in 2016 and has been routinely used with the Octavius 4D phantom for patient specific QA measurements. Over time an increasing number of ion chambers showed signs of leakage currents that could be seen as spikes during the acquisition of the standard 12x12 cm warmup field. When the detector is functioning normally this field is flat and hence any spikes can be discerned at this point whereas it is more difficult to distinguish this during measurements of patient plans which can be highly modulated.

Method A retrospective analysis of PSQA results from 2016 until early 2020 was performed. 315 6X 12x12 cm warmup fields were located for this period. The cumulative dose for each ion chamber is stored as a function of time at 0.2 s intervals during the acquisition. A Python script was written to read in these files and analyse the response of the ion chambers. The ion chamber readings during the beam on time and immediately after were investigated. Results Figure 1a shows the cumulative ion chamber readings for the SRS1000 array recorded during a 12x12 cm warmup field. The x and y coordinates correspond to the position of the ion chambers in the array and the height of the bars corresponds to the absorbed dose. Several spikes can be seen in the densely packed central region of the array. Figure 1b shows the instantaneous dose rate measured for one of the ion chambers that showed leakage both during the beam and post irradiation. Conclusion The deterioration in performance of the PTW Octavius SRA1000 detector array has been shown to be related to an increase in post-irradiation leakage. The current warmup fields combined with some data processing can effectively monitor the level of post-irradiation leakage in the detector array and can be used in a preventative maintenance program to ensure the validity of patient specific quality assurance measurements and the safe delivery of radiotherapy to patients. Introduction MRI-Linac components produce a Lorentz force on secondary electron paths that present unique challenges for radiotherapy dosimetry [1] . Particularly for an inline MRI-Linac configuration, secondary electrons produce an enhanced, forward peaked dose along the central axis, particularly at the entry surface region. Earlier studies performed on the Australian MRI-Linac conclude that 2 cm of bolus placed upstream to the treatment surface is suitable to reduce electron contamination [2] . The purpose of this work was to investigate an alternative method, off axis irradiation, whereby the primary x-ray beam is shifted laterally away from the isocentre and separates contaminant electrons from the treatment area; potentially providing superior dose uniformity [3] .

Method Measurements were performed on the Australian MRI-Linac which consists of a 1T MRI (Agilent, UK) unit coupled with a 6 MV flattening filter free (FFF) Linatron-MP (Varex, USA). Beam profiles using Gafchromic EBT3 film were irradiated perpendicular to the beam and positioned in a 30 9 30 cm 2 solid water phantom at SSD=1.8 m. Point dose measurements were also taken with the microDiamond 60019 (PTW, Germany) and MOSkin TM detector. Three field sizes of size 2.1 9 1.9 cm 2 , 6.1 9 5.8 cm 2 and 10.1 9 9.7 cm 2 were investigated and shifted approximately 10 cm laterally from isocentre.

Results The strong magnetic field of the MRI unit causes secondary electrons to focus at the isocentre, as shown in Figure 1 . Contamination electrons result in a near surface dose of 118%, relative to d max as shown in Figure 2 which is readily absorbed by at least 20 mm of solid water. MOSkin TM measurements at 1 mm depth recorded 120% relative to d max .

Conclusion By shifting the primary beam laterally from isocentre, electron contamination can be removed from the treatment beam. The clinical implication of off axis irradiation is that standard radiotherapy beams can be achieved in a high field MRI-Linac without risking excessive patient skin dose.

O067 Audit development for online adaptive radiotherapy for MRI and CBCT based systems Introduction Patient anatomy varies during treatment due to physiological processes and treatment response. Cone-beam CTs (CBCTs) are acquired for patient setup but can be used to evaluate anatomical changes and plan adaptation needs. CBCTs suffer from variation in Hounsfield Units (HU) and reduced field of view, limiting dose calculation accuracy. We evaluate deformable image registration (DIR) to map HUs from the planning CT (pCT) to facilitate dose calculation on the CBCT for adaptive decision making. Method For 10 head and neck cancer patients with radical intent, a replan-CT was taken within an average of 6 days of a CBCT that triggered re-planning. DIR was performed between the original pCT and the CBCT to obtain the pCT HUs with the CBCT anatomy (synthetic CT, sCT). A b-splines multi-resolution algorithm in Velocity AI software was used. The treatment plan was recalculated on the sCT retrospectively. The dose as calculated on the sCT and the replan-CT was compared using local dose-volume metrics for targets and organs-at-risk.

Results Comparisons of the sCT with other patient images are shown in Figure 1 . Difference between dose metrics for the sCT and replan-CT are shown in Figure 2 . The mean ± SE difference between the sCT and replan-CT metrics was 2.6% ± 0.9. The comparison showed violation of protocol metrics in 70 cases (out of total 160) for both sCTs and replan-CTs.

Conclusion We have shown that DIR has the potential to be used in the clinic for adaptive radiotherapy as a means of determining the need for re-planning to account for anatomical change (Figs. 1, 2) . 

Introduction Quantitative MRI (qMRI) has high potential in radiotherapy treatment planning (RTP), with abilities to enhance treatment response monitoring [1] . However, challenges in determining its accuracy and repeatability is limiting its clinical utility. Using standardized phantoms for such assessments has been identified as one of the keys to overcome these challenges [2] . The NIST system phantom can be used to measure qMRI relaxation times; including T1 and T2 (in ranges found within human body) [3] . This study utilised the system phantom to assess variations in MR-derived quantitative values over time.

Methods The phantom consisted of a T1 and T2 layer; each embedded with 14 spheres containing solutions with known NMR derived reference values. It was imaged on a 3T Siemens Magnetom Skyra MR-scanner. Variable Flip Angle (T1-VFA) and Inversion Recovery (T1-IR) sequences followed by nonlinear least squares fitting were used for T1 map generation. A spin-echo (multi-echo) (T2-SE-ME) sequence and mono-exponential fitting were used for T2 mapping. Region of interest (ROI) statistics were then extracted from each sphere, and the coefficient of variation (CoV) was calculated from five monthly measurements.

Results T1-IR generated less deviations from the NMR references compared to T1-VFA: Respectively, av. %difference was 5% vs. 53%. These latter measurements are known to be sensitive to field inhomogeneity's [1] [2] [3] [4] . Further, the CoV was lowest for T1-IR (\1%) compared to T1-VFA and T2-SE-ME methods (see Figure 1 ).

Conclusion Significant deviations between the measured and NMR derived T1 and T2 values were observed. Repeatability measurements were also found to fluctuate depending on the qMRI sequence utilized; T1-IR achieved the smallest variations between monthly measurements. Further longitudinal measurements are to be acquired for determination of the accuracy and repeatability of the scanner; required for effective clinical implementation of qMRI.

Introduction During patient treatments on the Elekta Unity MR linac contaminant electrons within the primary beam and electrons produced within the patient, that enter air around the patient, spiral along the magnetic field direction [1, 2] . In this work we present early results of investigations to evaluate these two sources of out-of-field dose.

Method Following the methodology of Hackett et al [1] EBT3 film was used to determine dose due to spiralling contaminant electrons (SCE). Out-of-field dose in a B = 0 T environment (background dose) was compared to SCE dose in order to determine a 'background' corrected SCE dose estimate. A phantom was used to investigate out-of-field dose arising from electrons ejected from, and produced within, the phantom by an external beam. The phantom is aligned so that electrons ejected at the beam entrance and exit spiral towards a strip of EBT3 film. Dose profiles were determined from the film to quantify the electron streaming effect (ESE) [2] at the beam entrance and exit. The attenuation of electron streams was also determined. Finally, ESE due to beams traversing the anterior MR imaging coil was investigated. Results Out-of-field dose with B= 0 T is comparable to that associated with SCE, B = 1.5 T. A background corrected SCE suggests a reduced magnitude for the SCE effect. Electron streaming from the phantom surface at the beam entrance is less than that at the exit. Electron streams are effectively absorbed by 1 cm bolus. The anterior coil generates electron streams depending on the field size and coil inclination. Conclusion Doses due to the ESE are greater than SCE. Electron streams are effectively absorbed by 1 cm bolus. ESE from the anterior Figure 1 Repeatability of qMRI values obtained for each ROI Introduction This abstract details the experiences learned from commissioning a HexaPOD evo RT System on an Elekta Versa HD and highlights an issue that the commissioning physicist needs to be aware of. The HexaPOD system, with 6 degrees-of-freedom is highly beneficial to treatment delivery accuracy.

Method Numerous phantoms were used to assess pitch, yaw and roll of the HexaPOD system. These included the MIMI and thorax phantoms, Rando's head and a dosimetric assessment using the ArcCHECK (1). The HexaPOD and XVI systems were separated and tested individually. For HexaPOD, known translational and rotational shifts were passed off from the XVI Registration platform to iGuide and measured using steel rulers for translations, a digital inclinometer for pitch and roll, and trigonometry for yaw rotations. For XVI, bidirectional arcs were used to image an offset phantom, and to assess that the positional error shifts were the same for each arc direction.

Results In combination with the Elekta XVI kV imaging system, it was found that the HexaPOD could not achieve specification i.e. 0.2°r oll and sub 1 mm translations. However, when tested separately, by passing off known values from the XVI Registration platform, the HexaPOD performed to well within the specified tolerances. It was found that by performing bi-directional arcs, the XVI system recorded a difference in the (Y) co-ordinate (roll) of up to 0.8°. This was discovered to be a function of the XVI flex maps and a small shift in the kV panel as it rotated. Conclusion When commissioning a HexaPOD, it should be separated from the imaging system. In this study, it was found that the XVI coordinates provide the highest level of uncertainty in the system and need to be addressed. Before commissioning the XVI-HexaPOD combination the XVI needs to be fully characterised for rotational errors. Introduction Immobilisation devices are used to increase patient positioning reproducibility during radiation therapy. However such devices may result in an increased skin dose, reduced tumour dose and altered dose distribution. Our institute recently commissioned CDR Systems immobilisation devices which include several sets of the FreedomX total body immobilisation system, Sabella Flex breast immobilisation system and KOILIA MIKROS belly boards. The purpose of this study is to report the physics measurement performed for the commissioning of the new immobilisation devices prior to their clinical use.

Method Measurements using Varian Truebeam linear accelerator were performed to assess the transmission and surface dose enhancement of these immobilisation devices over a range of photon beam energies (6 MV, 6FFF, 10 MV, 10FFF and 18 MV) and square field sizes. Both cylindrical (transmission) and plane parallel (surface dose) ionisation chambers were used in solid water and Perspex phantoms. An end-to-end validation testing was also performed, starting from CT simulation to treatment delivery, following the departmental clinical workflow. Results Overall, 6 MV FFF beam energy showed the lowest transmission values for all three immobilisation systems. For the 10 cm 9 10 cm field size, transmission values of 0.96, 0.97 and 0.98 were measured for the belly board, the breast board and FreedomX total body system respectively. Similarly, relative increase of surface dose ranges from 26% to 60% from these devices for 10 cm 9 10 cm field size. The largest increase resulted from the belly board system for the 6 MV photon beam. The end to end test results were in excellent agreement with treatment planning system calculated dose (relative difference \ 1%) for VMAT type deliveries.

Conclusion The dosimetric properties of the newly purchased CDR systems immobilisation devices were characterised. It was found that these devices have been accurately modelled by the treatment planning system. Thus, these devices are deemed suitable for clinical use. Introduction This abstract details the characterisation of the SNC IC Profiler ion chamber array copper quad wedge for photons and the aluminium quad wedge for electrons for energy response [1, 2] .

Method Measurements of the energy sensitivity of the IC Profiler with the copper quad wedge were compared against a PTW thimble 0.6 cc chamber for the quantity D10% in plastic water. The photon energies assessed were 6 MV, 6 MV FFF, 10 MV FFF and 15 MV, delivered using a 30 9 30 cm 2 field size. Measurements of the energy sensitivity of the IC Profiler with the aluminium quad wedge were compared against a PTW Roos chamber for the quantity R50 in plastic water. The electron energies assessed were 6 MeV, 9 MeV, 12 MeV, 15 MeV and 18 MeV, using a 25 9 25 cm 2 applicator. Measurements over multiple days on two Elekta linacs, tracked the sensitivity and magnitude of change, by exploiting the daily subtle variations in beam energy.

Results Figure 1 shows the IC Profiler with the copper quad wedge tracking against the PTW 0.6 cc chamber for 6 MV D10%. Figure 2 shows the IC Profiler can track subtle photon energy changes with a comparable magnitude as a standard field chamber. Introduction Stereotactic radiosurgery (SRS) delivers small, focused beams, and its quality assurance requires measurements with high spatial and temporal resolutions [1] . The Octa is an innovative 2D dosimeter which provides temporally-resolved measurements, in real time and with a spatial resolution of 0.3 and 0.43 mm [2, 3] . This study was a first attempt at validating its use for quality assurance of SRS.

Method We considered a clinical VMAT plan to a cerebellar peduncle metastasis of size 0.422cc, delivered as a collapsed-beam at gantry and collimator zero. We constrained the delivery because the Octa was yet to be characterised away from this configuration. Centring was completed using a square field of 5 mm, with the Octa sandwiched between water-equivalent material at a depth of 6 cm, 94 cm SSD. Measurements were benchmarked against calculations with the treatment planning system (TPS, Monaco 5.11), compared in terms of relative dose on a point by point basis, within the 80%, and within the 80-20% dose level. Calculations were done on a dose grid of 1.0 mm, 1% uncertainty, but resampled on a 0.3 and 0.43 mm grid corresponding to the pitch of orthogonal and diagonal detector elements respectively.

Results Within the 80% level, there was a maximum percentage difference of -2.1% in the cross-plane profile and a maximum of -2.6% in the diagonals. Within the 80-20% level, there was a maximum percentage difference of -3.5% in the cross-plane. Cross-plane profiles with the Octa had a systematically lower intensity through the penumbra when compared to TPS (Figure 1, Figure 2 ). Likely Introduction A challenge in the material extrusion printing of bonelike densities based on clinical Computed Tomography (CT) is the limited control of observed Hounsfield (HU) numbers over 0 HU (water) when using commercially available filaments. This work developed a novel method to additively manufacture a range of densities above 1.0 g/cc using the interlaced deposition of two commercially available filaments -standard Polylactic Acid (PLA) and iron reinforced PLA (Fe-PLA). Method Thin PLA and Fe-PLA layers were interlaced for a total of 11 cylindrical test (L1-L11) specimens with different volume ratios, V ratio of Fe-PLA to standard PLA ranging from 0.1 to 0.9 (0 = solid standard PLA, 1 = solid Fe-PLA), as a function of layer thickness (mm) (see Fig. 1 ). Standard protocol of 140 kVp CT at 2 mm slice thickness and increments was used for CT scanning, and mean HU and standard deviations were measured from the CT-scans. Results The interlace deposition of cylindrical test specimens were successfully printed with minimal printing artefacts (oozing and stringing). Solid test specimens for standard PLA (V ratio of 0) and Fe-PLA (V ratio of 1) were found to have mean HU values of 84±23 and 2244±90, respectively. Furthermore, the L1 to L11 test specimens with V ratio ranging from 0.1 to 0.9 (Fe-PLA to standard PLA) were observed with mean HU values ranging from 281±12 to 1900±46 at 140 kVp (2 mm slice thickness) (see Fig. 2 ). Conclusion This work demonstrates a novel and effective method to additively manufacture the full range of bone-equivalent HU values for CT imaging studies, which have great potential for multiple applications in diagnostic imaging as well as radiation therapy.

Introduction Radiomics have shown promise for predicting outcomes in patients receiving radiotherapy for primary lung cancer. Often, however, patients are planned using motion-averaged or maximum-intensity projection reconstructions from 4D CT images which help to characterise extent of tumour motion but may degrade the appearance of highly-predictive radiomics features. In this study, Method 3D-printed lung inserts were created using repeating geometric structures (with modified fill repeat rate and fill factors from thermoplastic (PLA) as shown in Figure 1 . CT images were acquired as static (3D) and in 4D with motion ranges from 1 to 30 millimetres. Static views were repeated to determine baseline feature stability by test-retest. The threshold for stability for each of the 110 features was then compared to the change in radiomics signature on MIP and average reconstructions at each scanned motion range (1, 2, 5, 10, 20 & 30 mm). This was used to infer the reliable range of motion at which point any individual feature becomes unstable. Results Of the original 110 features, 86 maintained a stable value compared to natural test-retest variability for motion ranges less than 10 mm. Some inserts exhibited consistently better stability than others with motion. The majority of features changed from baseline sequentially with increasing motion, however nine were shown to be variable with motion indicating poor stability for 4D assessment. The majority of features were more reliable on maximum intensity projection reconstruction compared to the average reconstruction from all motion phases. An indication of degradation with motion for each insert and range is shown in Figure 2 . Conclusion This study has shown that the variability in radiomics values calculated on CT images from different motion ranges from 1 to 30 millimetres can be comparable in motion to the variability found in lung patients CT images. O079 A 3D printed phantom to assess geometric distortion of MRI scanners used in a radiotherapy setting Introduction Magnetic Resonance (MR) Imaging has become a standard imaging modality for target volume delineation in radiation oncology. System-induced distortion affects geometric accuracy, and thus can influence tumor defintion and the dose delivered to the target volume. In this study, we describe a novel phantom designed to assess geometric distortion of MR images in a radiotherapy setting.

Method The phantom contains a regular grid, and the external contour of the phantom is designed so it can simulate both head & neck and body patient geometry based on the phantom orientation. The phantom was 3D printed using an advanced selective laser sintering (SLS) 3D printing technology. A HP Jet Fusion 3D -4200 machine was used to print the device. The phantom featured an overall length of 308. 5 mm, a width of 246mm and a height of 264 mm including the feet and filler caps. The design also incorporates recesses inscribed on the surface of the phantom, suitable for alignment with the external laser systems available on MR simulators. The phantom can also be offset to simulate changes in geometry for different localisations of the target volume in the patient. The accuracy of the 3D print was assessed using a Toshiba Acquilion CT scanner as it has negligible geometric distortion. The phantom was scanned using 1.5T and 3T MR (Siemens) scanners for both T1 and T2 sequences to assess MR geometric distortion. Introduction Deformable Image Registration (DIR) has become a key tool for adaptive radiotherapy to account for inter-and intrafraction organ deformation [1, 2] . However, there is a lack of ground truth to validate DIR accuracy, particularly for low contrast regions [3, 4] . This work aims to develop a deformable phantom which can be integrated into an anthropomorphic motion platform as a multi-purpose QA tool, so that the deformable system can be used to quantitatively and objectively assess DIR performance.

Method Platinum cured silicone gel was mixed and added into a 3Dprinted tumour mould. Two-part expanding polyurethane solution (1 Part-A, 0.45 Part-B by weight) was mixed and poured into a cylindrical mould where it expanded. Inside the mould was the siliconetumour suspended in the middle. A CIRS motion phantom was modified so that the deformable lung-rod phantom with tumour and fiducial markers (FM) could be inserted into it. Lung-rod was compressed and scanned with 4DCT (Philips Brilliance wide-bore). The FM in different breathing phases were identified and mathematically erased from the images. The intact and edited images were deformably registered. Total registration error (TRE) and Dice Similarity Coefficient (DSC) were measured. Results The polyurethane-lung and silicone-tumour had a density around 0.1 g/cm 3 and 1 g/cm 3 and HU value around -920 HU and 115 HU. The lung rod phantom and silicone tumour showed elastic properties (Young's Modulus and Poisson's Ratio) in the same order of magnitude of literature [5] [6] [7] . The DSC for the tumour was above the recommend 0.8 value [2] , virtual FM co-ordinates were found to have a mean TRE between 1.8 mm-5 mm.

Conclusion This work developed a novel method to model deformable inserts using a foam and silicone phantoms in any shape, size and elasticity. Additionally, the use of fiducials and its digital erase [4] in the deformable insert could objectively measure how the algorithm performs in low-contrast areas.

O081 A comparison of CT-and photogrammetrygenerated 3D prints of a HDR brachytherapy surface applicator Introduction The introduction of 3D printing in Radiation Oncology has allowed for rapid and precise design of patient-specific surface applicators using patient CT imaging data. The use of photogrammetry for generating 3D printable surface applicators has several advantages over CT imaging including reducing the amount of ionizing radiation the patient is exposed to. In this study, we investigate whether an acceptable dosimetric plan can be obtained for a surface applicator designed using photogrammetry techniques and compare the plan quality to a conventional CT derived 3D printed applicator. Method The nose region of a humanoid RANDO phantom was the selected treatment site due to its complex topology. Photographic images were captured using a Nikon D5600 DSLR camera and reconstructed using Agisoft Metashape software. CT imaging data was obtained using an Aquillion LB scanner with 1 mm slice thickness helical scan and 120 kV p tube voltage. Virtual surface applicators with 10 catheter tunnels were designed in the software Blender and 3D printed with ABS plastic at 90 % infill. Treatment plans were generated and evaluated using dosimetric parameters while their fit assessed by air gap volume measurements.

Results Both surface applicators were printed with minimal defects and visually fitted well onto the RANDO phantom. Measured air gap volume between the photogrammetry applicator and phantom surface was 44 % larger than the CT applicator, indicating poorer conformity. The generated treatment plans were almost identical with a difference Phys Eng Sci Med in PTV V 100 of -0.83 % and skin D 0.01cc of 1.38 % being the largest discrepancies for the PTV and OAR, respectively. Conclusion A 3D printed surface applicator was successfully constructed for the RANDO phantom nose region using photogrammetry techniques. Although it produced significantly poorer conformity, its dosimetric plan was almost identical to that of the CT applicator, showing its potential for future surface applicator construction.

O082 A study of kilovoltage x-ray beam backscatter factors for different solid phantom materials Introduction The aim of this work was to create an automated workflow to allow radiochromic film (EBT3 / XD) to be used as an almost instantaneous dosimeter. Method Ten GafChromic EBT3 calibration films were exposed in 2 Gy increments up to 20 Gy. Calibrates were then scanned at fiveminute intervals post exposure over 24 hours using an AutoHotKey script resulting in 282 tiff images. Following the 24-hour scanning period, a MATLAB script was used to automatically read in the tiff images and create a series of 282 calibration curves distinct in time which is termed the ''temporal calibration model'' (TCM). The model is saved as a series of polynomial fit coefficients to net optical density as a function of dose, timestamped in five-minute increments. Five patient-specific film measurements were then carried out and scanned using the same five-minute scan intervals from five-minutes post exposure to 24 hours post exposure. The TCM was then automatically Gamma map comparison (2%/2 mm criteria) for a patientspecific QA measurement using EBT3 film (10 Gy max) scanned at *5 minutes post exposure and corrected using the TCM Phys Eng Sci Med applied using in-house software eFilmQA to convert the patientspecific QA films to dose based on applying the relevant calibration curve from the TCM corresponding to the arbitrary time post exposure that the film was scanned. Each dose plane at post-exposure scan intervals of 5, 10, 15, 20, 30 minutes and 24 hours was then compared to the treatment planning system dose prediction using gamma analysis.

Results The TCM accurately converted the film scan optical density to dose irrespective of the time of scanning post-exposure. Gamma pass rates normalised to the pass-rate after 24 hours post-exposure were found to be within 3% (3%/3 mm criteria) for all patientspecific QA measurements.

Conclusion We have developed a methodology that allows for radiochromic film to be accurately used as a dosimeter at any arbitrary scan time post-exposure, where previously a waiting period of 15 -24 hrs before readout was needed to ensure the post exposure development had stabilised. The creation of a TCM can therefore enable results from radiochromic film measurements to be obtained as quickly as needed (Figs. 1, 2 ).

O084 Cranial SRS phantom analytical methods used in dosimetry audits for complex treatments of multiple brain metastases.

Introduction We report the analytical methods used in assessing data from the ACDS end-to-end stereotactic radiosurgery (SRS) audit. The audit includes: MR imaging; small distal and dispersed targets; single or multiple isocentres; compatibility with various treatment platforms. Assessments must enable dosimetry credentialing in the TROG Local HER-O [1] and OUTRUN [2] trials, chiefly using film to report on the treatment's spatial accuracy, and diamond detector and film for dosimetry.

Method Facilities apply their preferred technique and clinical protocols in three audit cases and two audit cases have stationary beams. GTV/PTV and OAR dose constraints are defined. Targets are either delineated from facility MR images, of provided via DICOM RT structure set. An IMT MAX-HD TM (IMT, New York) cranial phantom with tissue equivalent materials and realistic bony anatomy suitable for IGRT is used. The phantom has CT/MR fiducial markers and cavities for dosimetry measurements. Dosimetry is performed using PTW microdiamond detectors (PTW Freiburg, Germany) and Gafchromic EBT3 and XD film (Ashland Inc., Bridgewater NJ, USA). Clinical immobilization must be used to hold the phantom. The resulting film planes in the phantom are often not congruent with 3D axes of the DICOM data.

Results Measurements have been conducted at six sites to-date. Point dose variations by microdiamond measurements range over approx. ±10%. Phantom film planes are tilted by up to 10 degrees with respect to the DICOM x-y-z planes. MatLab code has been developed to extract 2D dose maps at any tilt angle to allow assessment of the treatment at the film locations (Fig. 58) .

Conclusion Tilted dose plane extraction has enabled distance-toagreement and 2D gamma analysis metrics which are now being reported.

References/Acknowledgements Introduction Detectors used for small field dosimetry can exhibit field size dependent sensitivity [1] . It has been previously shown that EBT3 film requires no correction factor for small field dosimetry measurements in water due to its similar density [2, 3] . However to date, no studies have confirmed the suitability of EBT3 film for the conditions of extreme electronic disequilibrium presented by small fields measured within lung equivalent phantoms. This study used Monte Carlo simulations to evaluate EBT3 film in lung phantoms. Method All simulations were performed in DOSXYZnrc [4] . The phantom simulated was 20 cm in each dimension. In the beam direction it was layered 5 cm of water, 10 cm of lung (density of 0.26 g/cm 3 ) and 5 cm water -with each layer extending to the phantom edge. All simulations were performed with a 6 MV photon beam from a Varian iX linear accelerator for square fields sizes of side lengths from 10 mm to 30 mm. EBT3 film was simulated at various planes in the phantom (film perpendicular to beam direction), including at the centre of the lung and at the upstream and downstream interfaces between the lung medium and the water. For each film simulated, an identical simulation was performed where the film was replaced with lung. The correction factor required for film was calculated by dividing the simulation results of the lung phantom without film by the results for the film at each position.

Results EBT3 was found to over-respond by 2 to 5 % when the fields were very small [5] . Table 1 displays the central-axis correction factors required to account for this over-response. The uncertainty in each result is * 0.5 %.

Conclusion EBT3 film is an acceptable dosimeter to use without correction factors for 6 MV field sizes 20 9 20 mm 2 and larger. For smaller fields sizes, care should be taken to correct the over-response of EBT3 film in lung phantoms. References/Acknowledgements Introduction Using 3D normoxic polymer gel dosimetry techniques with MRI readout (1), this study presents a 3D dosimetric analysis of a 12th thoracic vertebrae (T12) VMAT SBRT treatment.

Method Gel dosimeters were manufactured using materials and methods described elsewhere [1] , poured into a 47mm 9 150 mm vial, simulating the spine and immersed within a lung ED tissue equivalent, water-soaked sponge wrapped in Superflab, simulating chest wall. CT images of the gel-phantom were exported to the TPS. A patient's bi-arc VMAT plan was copied onto the CT data set and recalculated retaining MLC beam shapes, creating a curved dose distribution around the T12 spinal canal (Fig. 1 ). The phantom with dosimeter was irradiated according to the plan and imaged using a 3T MRI with T 2 maps generated [2] . A 3D dose matrix was created from the individual dose maps and the 5 Gy isodose line of interest was plotted in 3D in MATLAB TM . The patient's CT dataset was imported into MATLAB TM and the spinal cord contour surrounded by the T12 vertebra was extracted as per Fig. 1 . From the fused images, the position of the vial that encompassed both the PTV and spinal cord contours was established. The key assessment criteria for this study was to observe if the 5 Gy isodose line was shaped around the spinal cord and the delivery did not violate the isodose constraint.

Results Figure 2 shows the 5 Gy (red) isodose sphere of interest wrapped around the spinal cord (blue). At no point does the 5 Gy isodose line violate the spinal cord isodose constraint. Conclusion This study found that the full delivery chain, from end-toend can safely deliver a high dose fractionation regime to complex and tight margins, resulting in improved patient outcomes. References Introduction This study improves upon a previous investigation of lung tumour peripheral doses using a novel in-house manufactured lung-tissue simulator [1, 2] . Method To simulate lung tumour, foam material was cut and shaped into the size of a human lung and wrapped with 1 cm thick Superflab bolus simulating chest wall. A 47 mm diameter hole was drilled into the centre of the sponge where a gelatine filled PET vial was placed.

The foam was saturated in water to be equivalent to that of average lung tissue electron density (*RED 0.250). The phantom was CT scanned and images sent to the MonacoÒ TPS for planning. Conformal arc plans were generated for 6 MV, 6 MV FFF and 10 MV FFF, with field size 4.8 cm 2 and 450 MU. The central axis transverse dose planes were exported for comparison with measurements. Gel dosimeters were manufactured using materials and methods described elsewhere [3] , poured into vials and immersed within the sponge. The sponge with dosimeters were irradiated according to the treatment plan and subsequently imaged using a 3T MRI with T 2 maps generated [4] . Profile comparisons between the TPS and dosimeters were performed using Matlab TM . Results Measurements show that MonacoÒ over estimates dose in the build-up peripheral region of tumour equivalent tissue in lung for the photon energies investigated, as per Figure 2 . These results are similar to results from an earlier version of this study [1] (Fig. 61) .

Introduction In the 1960s it was observed that irradiated alanine had a long lived radical that could be easily measured with EPR spectroscopy and the alanine/electron paramagnetic resonance (EPR) system was proposed as a tool for radiation dosimetry [1] . Currently, NIST in the US [2] the NPL in the UK [3] , and the Helmholtz Institute in Germany [4] have implemented an alanine/EPR dosimetry service. The Princess Alexandra Hospital recently purchased a Magnettech MS5000 EPR spectrometer as a research tool with the aim of developing protocols for EPR based in-vivo dosimetry for use in radiotherapy. Method A protocol for commissioning EPR spectrometers for radiation dosimetry was developed. Standard tests were performed that included, temperature and energy dependence, dose linearity, and reproducibility. Measurements were made using a calibration set of dosimeters that were irradiated with 2 MeV electrons by the manufacturer and with dosimeter pellets irradiated at the energies used clinically in the department. In addition, tests were made to control for possible sources of magnetic interference from linear accelerators in adjacent bunkers. Results Figure 1 show the EPR spectrum for the calibration dosimeter set with 2 MeV irradiation from 0 to 102 Gy. The EPR signal strength was linear with dose over the measured range. Conclusion The dose linearity and energy dependence are both suitable for dosimetry in a radiotherapy setting. As expected, a strong dependence of the EPR signal on measurement temperature was observed which can be controlled for by allowing the spectrometer to reach equilibrium before measurement. The largest source of error was in sample positioning which needs to be addressed in order to achieve the accuracy necessary for in-vivo dosimetry. References Introduction Reactive oxygen and nitrogen species (RONS) are active agents in radiation therapy, responsible for a large fraction of the cancer cell death. However in conventional radiotherapy, external beams deliver an entrance and exit dose to normal tissue, while the use of radioactive isotopes in brachytherapy has inherent handling risks. Cold atmospheric plasma has been shown to generate RONS when directly applied to cells in culture and when applied to the culture medium itself, potentially enabling the delivery of active agents without the constraint of radiation damage to normal tissue. The aim of this presentation is to summarise the potential of cold plasma to treat certain cancers and to outline the proposed instrumentation for in vitro treatment to verify the efficacy, before progressing to clinical trials. Method The cell death for cell lines cultured in vitro and exposed to radiation and to direct plasma was compared (Figure 1 , adapted from [1] ). The concentration of RONS caused by radiation therapy and by plasma exposure was determined by simulation with input from the literature. The distribution of RONS created by the plasma exposure was calculated from Brownian diffusion. The information concerning equivalent doses was used to formulate a new design of a plasma system to be used in in vitro studies and to support potential use in clinical trials. Results Our new plasma treatment system design is intended for use in a hospital based laboratory. It has two applications: the first is the exposure of cell lines for preclinical in vitro studies and the second is the exposure of biologically compatible solutions to enable pilot clinical trials. Conclusion Cold plasma treatment offers new opportunities as a potential replacement or supplement for existing cancer therapies. Figure 1 For a range of breast cancer cell lines the dose required to reduce the growth rate by 50% for exposure to radiation (Gy) and plasma in jet mode (seconds). Each point represents a single type of breast cancer (cell line). Adapted from [1] Introduction Acoustofluidic manipulation of drug vehicles such as microbubbles and echogenic liposomes ELIP, has promising theranostic applications. A new acoustofluidic delivery agent that has an extended echogenic and number concentration stability [1, 2] , has shown promising acoustofluidic enhanced intravitreal drug delivery potential in an ex vivo bovine eye [3] . The shell stiffness S p and shell friction S f coefficients of the agent are investigated as a prerequisite to studying ELIP translation due to sonication. Method Frequency dependent attenuation measurement of an acoustic beam through a suspension of liposomes at room temperature and at 37°C were performed under constant 11 kPa pressure excitation to avoid influences of frequency dependent transducer performance and non-linear bubble deformations. The acquired data were analysed using an established theoretical model described in details in [4] , to calculate the shell parameters. The agents were formulated as described in [5] , and their size distribution and concentration were measured using Tuneable Resistive Pulse Sensing as in [2] . The bubbles solution was then diluted by 1:20 and placed in a test chamber between an ultrasonic transducer and hydrophone, Figure 1 . The transducer was excited using an arbitrary function generator at a frequency range of 2-20 MHz and the pressure after passing the bubbles suspension was acquired. A peristaltic pump circulated the bubbles in the test chamber between tests to compensate for buoyancy effects. Results The frequency dependent attenuation experimental results and the best fit are shown in Figure 2 . The new bilayer acoustofluidic delivery agent shell parameters were calculated as (°C) and (°C). Conclusion The new echogenic liposome shell parameters calculated in this study show very low stiffness and friction coefficients despite their extended stability, which makes them unique agents for acoustofluidic delivery applications.

Introduction During the outbreak of COVID-19 in 2020, there is a high demand of isolation rooms in hospitals for patients. Considering that it is technically not feasible for conducting major modification of ventilation system in hospitals within a short period of time, a fasttrack and cost-effective transformation approach was developed to convert general ward to negative pressure ward in public hospitals in Hong Kong to cope with the upsurge in suspected cases of COVID-19. Method A Mobile Modular High-Efficiency Particulate Air Filter Unit (MMHU) was developed to transform a general ward to a negative pressure ward quickly and with low-cost. The MMHU is a tailor-made mobile air purification unit with High-Efficiency Particulate Air (HEPA) filter and resembles a large air cleaner with exhaust fan and HEPA filter to extract indoor air from ward area through opening in window to outdoor, thereby creating a negative pressure inside the ward. Results MMHUs were deployed to public hospitals and field tests including pressure difference measurement and smoke test were conducted. The results proved that a general ward transformed with MMHUs met the reference parameters advised by the Guidelines for Environmental Infection Control in Health-care facilities, and the use of MMHU created an inward flow from the corridor to the ward, thus preventing cross-infection inside the ward. The use of MMHU could minimize the major modification of ventilation system with traditional means of conversion and shorten the conversion time. It also helps to ease public concern on the air exhausted from wards. Conclusion The use of MMHU is able to transform a general ward to a negative pressure ward within a short period of time and such transformation is considered effective in increasing the isolation facilities capacity in hospitals during the outbreak of infectious diseases. The COVID pandemic has highlighted the fact that clinical engineering has evolved from being management of instrumentation maintenance to playing a significant role in the medical team and the health system.

Training beyond basic undergraduate studies is necessary to perform this role competently and effectively: among practising Clinical Engineers there has been an increasing awareness of the need to upskill to manage changes in technology.

Our experience supervising working biomedical engineers completing part-time higher degrees in research confirms our view that optimal biomedical engineering education programs need to be a full five years in duration.

Typical projects involve part-time higher degree by research students working on clinical projects hosted by hospitals/ research establishments, in which some of the candidates are also performing their normal work.

Informal associations with clinicians and requests for collaboration in projects are an important part of this process.

Currently in Australia there are more than ten biomedical engineering university degree courses [1] .

To remain on par with the large number of graduates being produced from these courses, experienced Clinical Engineers practising in the workforce are seeking to upgrade their qualifications to postgraduate level.

This paper discusses a number of higher degree projects that illustrate these points. Reference 1. https://www.engineersaustralia.org.au/About-Us/Accreditation P004 Engaging with a rapidly changing risk environment: how the Australian Clinical Dosimetry Service responded to CoVID-19

Introduction On the 17th March, with an unknown potential of a SARS-like virus beginning to impact Australia and a logarithmic increase in new cases nationally, the ACDS team was instructed to work from home until further notice. Five days later on Sunday 22, all audits were grounded until further notice. This included a Level Ib audit which was scheduled the following day on a newly installed linac. The ACDS, in collaboration with subscribers and regulators, has established new protocols enabling safe auditing to recommence. Method Operationally, facilities scheduled for on-site Level II and III audits were notified two weeks prior that they would be re-scheduled and that the conditions of audit would be modified to minimise the likelihood of infection. The Level I mail-out audit for Qtr 1 was being returned for read-out during this time. Returned audit kits were disinfected prior to being picked up and taken to the responsible staff member's home, where the readout equipment had been relocated. Strategically, the jurisdictional regulators and the Department of Health Division responsible for managing the Radiation Oncology Health Program Grants (ROHPG) were contacted and the CoVID-19 impact was discussed.

Regulators recognised the necessity of auditing to ensure patient safety and all agreed that they would support recommencing audits in a controlled ramp-up. DoH recognised that CoVID-19 would affect the audit schedule, and would monitor developments.

Results An entire new set of procedures for auditing staff was developed and for the first time, some audits were performed with remote support. The first audit post grounding was on the 12 April in a jurisdiction with closed borders.

Audit teams of two ACDS staff were assembled to work together and no team performs more than one audit trip a week. Conclusion The new normal is a constant change. The ACDS must respond quickly and appropriately to fulfil its important role in Australian radiation safety.

P006 Modelling radiobiological response of paediatric cranial tumours following intensity modulated proton and photon radiation therapies -radiosensitivity and gender dependence Introduction Medulloblastoma is the most diagnosed paediatric malignancy of the central nervous system. Despite the high number of paediatric patients treated with intensity-modulated photon (IMRT) and proton (IMPT) radiotherapy, risk factors for increased normal tissue complication probability (NTCP) are still not fully understood. Disparities in intrinsic radiosensitivity (a/b ratio) for paediatric patients exist, influencing NTCP outcomes. Additionally, literature suggests that TD50 can be *20% lower in female patients. Following this data, our aim was to model the impact of gender and intrinsic radiosensitivity on paediatric brain tumour IMRT and IMPT outcomes. Method 216 comparative IMRT and IMPT plans were created for six paediatric patients, using Varian Eclipse software. Relative-seriality and Lyman-Kutcher-Burman models were used to calculate NTCP values for cranial irradiation of several anatomical structures.

Graph 1 Effect of TD50 on NTCP of brainstem necrosis and cochlea tinnitus for medulloblastoma patients. TD50 for females is approximately 20% lower than males, this is represented dashed line compare to the second line (underlined TD50) which illustrates the mean Therefore, females will potentially have a higher NTCP for the given dose distribution Phys Eng Sci Med Evaluated complication endpoints included brainstem necrosis, blindness, and deafness. Sensitivity analyses were performed to gauge the impact of a/b values and TD50 on radiation response. Results IMPT plans demonstrated smaller side-effect risks compared to IMRT across all NTCP models. For medulloblastoma, tinnitus and brainstem necrosis NTCP depended on modality and TD50 of irradiated normal tissue, therefore NTCP could be potentially underestimated in female paediatric patients (Graph 1). Similarly, IMPT demonstrated advantages for NTCP across all tested a/b values compared to IMRT, with lower a/b values estimating a reduced risk of adverse side effects (Graph 2). If the true a/b of these paediatric tissues are higher than predicted (a/b *3), planning could severely increase risk of cochlea tinnitus for IMRT. In general, NTCP for IMPT was not as influenced when changing parameters in the model compared to IMRT. Conclusion This study aimed to estimate the impact of gender on NTCP for paediatric cranial critical structures. Modified NTCP calculations can be used for ranking of treatment plans to better stratify patients who would benefit most from IMPT.

Introduction The microenvironment of many solid tumours can be acidic, reaching an average extracellular pH value of approximately 6-7.0 [1] . Tumor pH is physiologically important since it influences a number of processes relevant to metastasis and treatment management [2] . Proton magnetic resonance spectroscopy ( 1 H MRS) and chemical exchange saturation transfer (CEST) are two sensitive techniques that provide metabolic information. The aim of this study was to evaluate the feasibility of pH measurement in vitro using 1 H MRS and glucoCEST. Method 1 H MRS and glucoCEST were performed on a 3T MR scanner. Five phantoms containing 20 mM of glucose were prepared at different pH values 5.4, 6.0, 6.5, 7.1 and 7.8. Model solutions were buffered using phosphate-buffered saline (PBS). Sodium-azide was used to suppress the growth of gram-negative organisms. All model solutions were pipetted from highly concentrated stock solutions to ensure a precise adjustment of the concentrations, and pH values were adjusted to the final values using hydrochloric-acid (HCl) at 92% of the final volume, to take into account the added amounts of HCl. Final pH values were checked by means of a calibrated pH electrode. Results There was no significant difference in resonances of glucose signals derived from 1 H MRSI for different pH (p[0.05) (Figure 1 ). For CEST imaging, the field inhomogeneity correction was performed by interpolating and shifting the Z-spectra based on the obtained B 0 Graph 2 Effect of a/b on NTCP of brainstem necrosis and cochlea tinnitus for medulloblastoma patients Phys Eng Sci Med field map. The optimal glucoCEST contrast obtained with RF power of 3 lT in the current study ( Figure 2 ). Glucose peaks in CEST Z-spectra of the phantoms were observed between 1.1 ppm and 2.9 ppm. The CEST contrast showed a significant negative correlation with pH (r = -0.87 and p \ 0.01). Conclusion These results demonstrated the potential role of glu-coCEST for microenvironment pH mapping in vivo. Acquisition parameters optimization is necessary to obtain the maximum glu-coCEST imaging contrast. References Introduction The ring applicator is used for high dose rate afterloaded brachytherapy of the uterine cervix. The Iridium high dose rate source is welded to a cable that is unwound by precise amounts by a step motor to control the placement of the source in the applicator.

Being circular with a small radius and narrow lumen, the ring modifies source displacement. To predict the location of source for accurate dose calculation a factory source path is used. The aim of this study was to characterise the reproducibility of the placement of the source for a ring applicator and determine the accuracy of the factory source path. Method The ring applicator was fixed to EBT3 film in a rigid assembly and radiographed using the superficial unit to deliver exposure adequate for visualizing the internal structure. The rigid assembly was transferred to the brachytherapy suite and autoradiographed with the Iridium source at nominal dwell positions spanning the circumference of the applicator. Image analysis was conducted using Excel spreadsheets and commercial film analysis software. Several repetitions of the process were performed for assessment of reproducibility of the source location. Introduction It has been noted that that when using the ''Dose to water'' option of a voxel based TPS to calculate a depth dose graph through a HD block there is a rapid rise in dose at the boundary of normal density water to a high-density block of bone like material. This rapid rise has been considered not realistic. This paper presents Conclusion The standard interpretation of radiation interacting with a dosimeter is supported. That is, the material of the dosimeter responds to the radiation fluence. The radiation fluence is greater in the HD bone like material than in the water, hence greater response. The active medium of GafChromic film is designed to be equivalent to water, therefore the response of the film being in agreement with the ''Dose to water'' calculation of the TPS is as expected. Introduction For MV radiation, it is simple to demonstrate that for a ''Dose to medium'' calculation of a TPS using a small voxel size, less than range of high energy electrons, that if the ''medium'' of two adjacent voxels are different, then the value of dose derived for these two voxels will be different by the MSPR of the two materials that constitute the ''medium'' of the respective voxels. This in itself is a trivial conclusion, however applying this principle to real patient plans where there are one or more regions of significantly different material within a PTV, then it can be concluded that for certain patients, it will not be possible to delivery a homogeneous ''Dose to Medium'' distribution.

For different TPS and different optimisation criteria, different consequences of this effect can be observed. However, for all EBRT, optimization is the adjustment of the primary beam through the collimator (MLC). All TPS optimisations using ''Dose to Medium'' calculation will exhibit unstable behavior when they encounter heterogeneous material distributions, in particular for clinical patients, bone to soft tissue, within the PTV (Fig. 72) .

For the Elekta Monaco TPS, this may result in the optimization not being able to achieve an acceptable result (fail), or exhibited as an overall increase in total radiation exposure to the patient. That is, an increase in radiation to uninvolved tissue when ''Dose to Medium'' is used compared to an equivalent plan optimised using ''Dose to water''. Method Patient plans were developed on the Elekta Monaco TPS using both ''Dose to Medium'' and ''Dose to Water'' options.

Conclusion It is concluded, that it is a problem to use ''Dose to Medium'' calculation for optimization. It can be unstable and overall, always results in a greater total radiation exposure to the patient. Introduction Incomplete charge collection due to the ion recombination, P ion , in ionisation chamber requires a correction factor for accurate dose measurements. Generally, three main factors affecting the P ion in a pulsed Linac's beam which are: applied polarising chamber voltage, plate separation and the dose per pulse (DPP). The purpose of this study was to investigate the P ion correction factors for three different ionisation chamber models under conventional flattened beams as well as in FFF beams using three methods. Methods P ion factors were measured using Jaffe' plots, two-voltage method and a semi-empirical method derived from the Boag theory 1, 2 which utilises the dose per pulse from the linac for three PTW type chambers: Farmer TW30013, Semi-flex TW31010 and Pinpoint TW31016. Measurements were performed using a 10x10 cm 2 field size at 100 SSD and at 10 cm depth on the CAX for 6 MV, 18 MV, and 6FFF beams produced in a Varian Clinac IX. The measured P ion factors were compared and validated against the published guidelines 2 . Results P ion factors measured for the three ionisation chambers were found to be generally agreed with the published data 1, 2 . The twovoltage and a semi-empirical methods indicated similar results. However, the Jaffe' plot showed variations of maximum value of 0.13 % from the other two-methods. It was also evident that P ion factors were slightly varied for flattened beam and FFF beam among three chambers and this is most likely due to the effect of dose per pulse for different beams and detector size variations. Conclusion It is necessary for all clinical departments to characterize their ionisation chambers before being utilized for clinical dosimetry. The study demonstrated that the two-voltage method can be considered as a suitable method for determining P ion factors for FFF beams. A series of D-a-alanine dosimeters were irradiated in intervals of 7 Gy up to 84 Gy using 6 MV photons. Four separate EPR signals were acquired for the capsules/tablets using two orientations (up/down) and two rotations (0/180°). Alanine EPR spectra were analysed using the double integral, peak-to-peak height and gradient of the main absorption peak to compare each technique for accuracy as an estimator of dose.

Results The initial dose uncertainty for the Standard L-Alanine pellets and D-Alanine capsules was found to be 3.9 Gy and 4.5 Gy, respectively using the double integral analysis. Using progressive signal accumulation for doses less than 10 Gy these uncertainties for the standard pellets were reduced to 3.6 Gy using the double integral, 2.6 Gy using peak-to-peak and 2.5 Gy using the signal gradient method.

Conclusion The use of alanine dosimetry could be commissioned for clinical use provided the dose uncertainty is able to be minimised through either the use of a reference sample to correct for fluctuations in the spectrometer g-value. Alternatively, multiple reference dosimeters could be measured to correct the calibration curve for the g-value at the time of measurement. The dose should be measured using peak-to-peak values using progressively higher signal accumulation time for lower expected doses. The positioning of the dosimeter in the spectrometer sample holder produces a large uncertainty which can be minimised by fixing the dosimeter position within the spectrometer.

The author would like the thank Kaz Hosokawa, John Barry and Dr Sarah Walden from the Queensland university of Technology for their assistance in supplying some of the materials for this project.

Introduction The Princess Alexandra Hospital recently purchased a Lexsygsmart OSLD/TLD reader to provide in-vivo dosimetry within the department. The Lexsygsmart reader was supplied with BeO dosimeter chips and although they have good tissue equivalence and excellent energy independence they are not as well characterized as the more common Al 2 O 3 dosimeter chips. The current bleach cycle appears to result in a progressive reduction in the sensitivity of the dosimeters. A thorough understanding of the defect states involved in the OSL signal is vital for developing bleaching/annealing protocols that would provide sufficient accuracy for routine in-vivo dosimetry.

Method A BeO chip was first irradiated with 50 Gy using a 6 MV 10 9 10 photon beam. The EPR spectrum was measured using a Magnettech MS5000 spectrometer. The peak-to-peak amplitude of the Li?-hole EPR center was monitored together with the OSL signal while bleaching with blue light in the OSL reader. Initially the chip was bleached with 30 mW /cm 2 blue light for 30s. At the end of the bleaching periods the OSL signal was measured and the chip was then placed in the EPR spectrometer for measurement. This was repeated 5 times. The chip was then bleached with the standard protocol with 100 mW/cm 2 blue light for 300s. Finally, the chip was heated to 260°C for 300s. Results The EPR signal decreases in intensity progressively as the chip is bleached (Figure 1a) . The final spectrum after the full bleach cycle shows that the EPR centre has not been fully bleached at this point. This can be seen more clearly when the peak-to-peak intensity is plotted together with the OSL signal as a function of the number of bleach cycles (Figure1b). Both the EPR and the OSL signals decay but at different rates. Annealing at 260°C for 300s appears to have effectively annealed out the EPR defect. Conclusion These results show that the BeO dosimeters are not returned to their initial pre-irradiation state by the current bleach cycle. Approximately 30% of the EPR active defects produced by the irradiation remain after bleaching and this is related to the reduced sensitivity of the dosimeters after an irradiation/bleach cycle. P015 Implementation of advanced radiotherapy technology to improve clinical outcomes in regional set up Introduction In Australia, radiotherapy (alone or with chemotherapy) is used for treating 48.3% of notifiable cancer patients and availability of advanced radiotherapy would have better clinical outcome. Cancer patients in remote areas experience poorer outcomes than their metropolitan counterparts with distance being an appreciable barrier to treatment access mainly to advanced techniques. Problem At Central West Cancer Care Centre (CWCCC) the utilization rate of Intensity-Modulated Radiotherapy (IMRT) and Volumetric Modulated Arc Therapy (VMAT) was significantly lower compared to other NSW public health services. Stereotactic Ablative Body Radiotherapy (SABR) treatment was not available at CWCCC. Design Increase the IMRT/VMAT utilization rate and make SABR treatment available through quality improvement projects with multidisciplinary collaboration. A driver diagram (Figure 1 ) was developed to identify the challenges, drivers, change ideas and the aim of the project using recommended quality improvement tools from the clinical leadership program of clinical excellent commission of New South Wales. Key measures for improvement Increase IMRT/VMAT utilisation rate; SABR available to CWCCC patients. Strategy for change A multi-discipline team consists of Radiation Oncologists, Medical Physics Specialists and Radiation Therapists was formed and a project plan was developed to implement each treatment technique. What this study adds?

1. State of the art treatment can be implemented in rural settings with a proactive and collaborative team Figure 1 Combined EPR/OSL bleaching experiment a) The peak-topeak amplitude of the EPR spectrum decreases during optical bleaching but is not completely removed b) The OSL signal decreases much more rapidly than the EPR signal and is completely removed by the full bleach cycle • Additional staff and/or extensive training are not always required to implement new treatments • End-to-end testing to ensure safety and treatment quality is a critical part of practitioner confidence in new treatments.

Results After implementing Breast IMRT, VMAT and SABR treatment, the average IMRT/VMAT utilization rate at increases from 22% to 63% as shown in Figure 2 . SABR treatment is also available at CWCCC from 2019. The radiation oncology team is working to further maximise the utilization rate in coming years with the promising data of IMRT/VMAT utilization rate of above 70% as shown in Figure in January 2020.

Conclusion Strong commitment from a trained team and a collaborative approach is important for the implementation of advanced technology in regional centres.

Introduction The Varian Halcyon iterative cone beam CT (iCBCT) has previously been shown to be a suitable substitute for a diagnostic CT scanner for the treatment planning of Pelvis patients [1] . As Halcyon linacs have preset iCBCT modes that cannot be changed by the user, all kV settings are identical. Our institution has 4 Halcyon linacs, all of which have been commissioned for treatment planning scans. The aim of this study was to compare the data from each linac to determine the feasibility of using a single CT-ED curve for all Halcyons, reducing commissioning times, standardising QA processes, and reducing the risk of incidents involving incorrect CT-ED curves being assigned.

Method The CT-ED data of the Pelvis mode from each Halcyon linac was averaged to create a single CT-ED curve. The results from each linac were then compared to the average to determine how large the uncertainty would be in both CT-ED curves. The CT-ED curve for other predefined iCBCT modes was also collected and the differences compared to the curve determined.

Results The comparison of CT-ED curves for the preset Pelvis iCBCT modality across 4 Halcyons is shown in Figure 1 . The maximum difference between any datapoint on the CT-ED curve of any one linac was no more than 20 HU from the average, with the largest differences observed in lower density materials.

Conclusion This study has shown that there is potential for a ''standard'' CT-ED curve to be used for all Halcyon linacs. However, extending this beyond the Pelvis preset iCBCT mode may introduce some uncertainty. Introduction The advancement in imaging modalities has led to the accumulation of a large amount of data which requires significant time for contouring and processing the data for clinical use. It can also lead to interobserver variations and human errors in delineating target and organs at risk (OARs) volumes. In this study, an effort has been made to develop an automatic segmentation tool and tested on lung OARs.

Method A deep learning-based model known as U-net architecture was used to develop and train a model to auto contour the OARs including the spine, heart, and both lungs. The model was trained for each OAR with 62 datasets for lungs using 100 epochs, the spine had 98 datasets and 23 patients for the heart with 200 and 100 epochs respectively. Upon training, the models were tested on datasets and their results were evaluated based on the dice similarity coefficient (DSC) and Hausdorff distance (HD).

Results The average DSC of left and right lung was found to 0.96 and 0.95 respectively whereas the spine produced 0.86. With the lowest number of patients for heart, the model was able to yield the highest DSC of 0.97. In the case of evaluation using HD, the average distance between predicted and ground truth for the left lung was 3.0 ± 0.8 mm and was marginally higher for the right lung of 3.6 ± 0.70 mm. The lowest HD was observed in the spine, which was around 1 mm, whereas the heart model yielded an average distance between both contours of 3.52 ± 0.68 mm.

Conclusion The model demonstrates the potential to be used in a clinical environment given that thousands of patient data were used in training. However, this study is still under investigation and more factors will be considered to make it more user friendly, reliable, and accurate for clinical purposes.

Introduction The introduction of MRI linear accelerators (MR-linacs) requires improved approaches to MRI-only radiotherapy. MRI provides excellent soft tissue visualisation for target and organ definition, but no electron density information for dose calculation, obtained instead from registering CT images [1] . MRI-only radiotherapy would remove registration errors and reduce patient discomfort, workload and cost [2, 3] . Electron density requirements may be addressed in different ways, from manually applying bulk density corrections [2, 4] , to more computationally intensive methods, such as atlas based techniques [5] which automatically segment and apply bulk density corrections to produce substitute CT datasets (sCT). Reducing MRI imaging time would reduce potential artefacts from intrafraction motion and patient discomfort. This study investigated the effects of MRI imaging time reduction on sCT generation for prostate MRI-only treatment planning.

Method 10 volunteers were scanned on a Siemens Skyra 3T MRI. Sequences included the 3D T2-weighted (T2-w) SPACE used for sCT conversion as previously validated against CT [5] , along with variations to this sequence in repetition time (TR), turbo factor, and combination of these to reduce the imaging time. Additionally, a T1 DIXON scan was taken to test the robustness of the sCT technique. All scans were converted to sCT and evaluated for anatomical changes and dosimetric differences for a standard VMAT plan compared to the previously validated SPACE sequence.

Results Compared to the previously validated T2-w SPACE sequence [5] , scan times were reduced by up to 80%. The external volume and bony anatomy were compared, with most meeting a DICE coefficient of 0.9 or better, with the largest variations occurring at the edges of the external body volume. Both Octavius and Pinnacle TPS had been commissioned for zero-jaw asymmetric fields using water phantom scans and field output measurements. Water phantom scans in DIBH asymmetric conditions were not available for EPID commissioning purposes at the specific depth (1.5 cm) and SSD (98.5 cm).

Method EPID in conjunction with EPIQA software based on the GLAaS algorithm (1) and the Octavius chamber array together with PTW Verisoft software allows for Gamma analysis QA. TPS fluence maps were compared with both the EPID/EPIQA and Octavius/Verisoft using the Gamma Function (3%/v), with a pass/fail rate acceptance of 90%. While the EPID measurement depth is fixed at 1.5 cm, 3.7 cm measurement depth was chosen for the Octavius to maximize the Introduction AlignRT is a surface-guided radiation therapy (SGRT) technology that can monitor inter-and intra-fraction patient motion. Congruence of the SGRT and MV isocentres (''isocentre calibration'') is vital to prevent artificial walkout caused by rotation about different isocentres [1] . This is particularly true when using non-coplanar beams. Despite this, there is little research discussing the stability of isocentre calibration or suggestions for QA tolerances.

Since installing AlignRT at GenesisCare Southport we have performed pre-treatment QA of the AlignRT system for each singlefraction stereotactic procedure. These measurements provide information on the long-term stability and accuracy of the isocentre calibration. Method We performed pre-treatment QA using the AlignRT calibration cube in accordance with manufacturer recommendations [2] . AlignRT saves the isocentre calibration results as a report in PDF and XML formats. For this work we manually inspected the XML reports and copied the relevant information to a spreadsheet for analysis. Results We performed analysis on 53 measurements taken over 12 months from May 2019 to April 2020. Average discrepancies between the AlignRT and MV isocentres ranged from -0.06 mm to ?0.04 mm in the three translational directions. The maximum standard deviation was 0.09 mm. Rotational discrepancies had a maximum standard deviation of 0.06 degrees. We plot the congruence of AlignRT-MV isocentres over time in Figure 1 . No long term trends are visible. Conclusion We performed an analysis of pre-treatment SGRT QA results. The results show that the AlignRT-MV congruence calibration is stable and accurate. We suggest reducing the frequency of isocentre calibration QA to monthly. From the results of the analysis we suggest an action level of 0.2 mm or 0.2 degrees for the isocentre calibration.

Introduction Beam quality index (BQI) verification is part of the recommended periodic QA tests by the AAPM and IPEM. Common implementation from vendors utilise a set of wedges to determine the beam energy. Another common method is to use ratios of ionisation at two depths to verify beam quality constancy, which is time consuming. This study compares wedge implementations by Sun Nuclear Inc (Quad Wedge) and PTW Dosimetry (BQ -Check) for verifying constancy of BQI of photon beams and energy of electron beams. Method Quad-wedge phantoms were used with SNC IC Profiler array and SNC Profiler software. The BQ Check phantom was used with PTW Starcheck array and PTW Multicheck software. Copper and aluminium wedges are used for photon and electron beam measurements respectively. SNC quad-wedges requires a 30 cm 9 30 cm field for use with photons and a 25 cm applicator for use with electrons. The PTW BQ check cannot be used with FFF beams. The measurements were compared against reference PDD data measured with water-tank and PTW Roos chamber on the same day. Results Deviation in energy measurements compared to water-tank measurements are given below. The largest deviation was 1 mm in R50 for both PTW and SNC. Photon beam quality was within 0.5% and 0.3% for PTW and SNC respectively. Introduction Geometrically accurate delivery of LGK radiotherapy treatments is dependent on precise alignment of the PPS with the RFP. QA of PPS-RFP coincidence is performed with an Elekta diode tool and a film holder tool. Elekta also provides a SWP which includes a film insert. Film is firmly immobilized in the phantom with two rods inserted in holes generated with a specialized hole punch. Here we investigate the reproducibility of these hole-punch alignment marker positions, with a view to generating a third independent PPS-RFP coincidence test, as well as an efficient and accurate tool for aligning film-measured and TPS-calculated dose-maps. Method QA following a recent service included acquisition of two SWP irradiations each of the 4 mm, 8 mm, and 16 mm collimators with EBT3-film aligned in the axial plane. The digitized (400 dpi) films were analysed for reproducibility of the hole-punch-measured PPS isocentre within the pairs of film, and the difference in measured RFP of the three collimator irradiations. Analysis was performed in Python. The centre of each hole was measured by generating a binary-image using a global threshold of the blue-channel image, followed by morphological-closing and opening to remove noise. A region-growing-algorithm separated the hole-regions. PPS isocentre was calculated as the mean pixel position of the two-hole-regions. RFP was similarly measured. Results As summarized in Table 1 , the measured PPS-RFS was reproducible within pairs of axial-digitized-films (max discrepancy of 0.021 mm). The mean shift for each pair (column ''Film PPS-correction'' in Table 1 ) provides the shift to be implemented when performing PPS-vs-RFP QA using this method. Conclusion Although more repeat-measurements (including of coronal-oriented film) are required to verify these results, an additional PPS-vs-RPS coincidence QA-method has been developed. Introduction The machine performance check (MPC) is an integrated quality assurance (QA) tool on a Varian Halcyon linear accelerator (Halcyon). The aim of this study was to evaluate its suitability and reliability for daily output and uniformity checks. Method Measurements were done on a Halcyon located at the Icon Cancer Centre, Toowoomba, over 28 months, for the 6 MV flattening filter free photons (6 MV FFF) beam. During this time the Halcyon has undergone an upgrade to Version 2.0. MPC uses the existing megavoltage electronic portal imaging device (EPID) to assess the beam output and beam uniformity variation. The MPC uniformity measurement variations are given as a single result that comprises both vertical and horizontal planes.

MPC checks were compared against the daily checks acquired with the Sun Nuclear Daily QA3 device, as well as the monthly checks using PTW Farmer ionization chamber (IC) for output measurements and a PTW Octavius 1500 detector array (1405 planeparallel vented ionization chambers) for symmetry checks. Results The MPC-measured beam output drifted by 1.46 % per year comparing to the monthly Farmer IC and Daily QA3 measurements, taking an average of 4 months to change by 0.5%. MPC uniformity results were within 1.0 % from the monthly, and within 1.5 % from the daily symmetry measurements for Halcyon Version 1. Following upgrade to Version 2, MPC uniformity was within 0.5 % and 1.0 % from the monthly and daily symmetry assessment methods, respectively. Conclusion MPC on Halcyon was verified as a daily QA tool for Halcyon beam output and symmetry against independent detectors over 28 months. Trends show it's reliable and stable, observed drift in MPC-measured output was slow in comparison with the frequency of QA using the Farmer IC. The absolute difference is subject to the reference values acquired. Introduction Patient specific QA (PSQA) is an important part of the treatment pathway to ensure that the treatments delivered to the patient match what is calculated in the planning system. For Stereotactic Body Radiation Therapy (SBRT) type treatments, small fields and high levels of modulation can create larger discrepancies between measurements and planning systems. Monte Carlo (MC) offers a third gold-standard system to help determine if the differences are due to the planning system calculations, or the machine delivery. This investigation has implemented a Monte Carlo system using EGSnrc/ BEAMnrc/DOSXYZnrc [1] , MCtools [2] and custom in house code to allow the regular use of MC simulations to be compared with measured and calculated data. Method VMAT SBRT Arcs were simulated using BEAMnrc in the PSQA Geometry (dMax, 98.5 cm SSD) used by EPIQA to allow comparison against the Pinnacle treatment planning system (Philips, Netherlands) and measured EPID data. EPIQA (Epidos, Slovakia) uses an integrated EPID image acquired during the arc which is then converted to dose using the GlAaS Algorithm [3] . MC simulations were performed to a custom water phantom that matched the Pinnacle configuration, allowing the export and comparison of DICOM dose data in the EPIQA software. Brain and Lung treatments were simulated with a total of 19 arcs simulated and compared with measurements and Pinnacle calculations. Results Simulations took in the order of 3-5 minutes per arc to achieve an uncertainty of .4% on a dedicated compute cluster with 192 threads. For the Brain treatments all arcs passed a gamma analysis at 2%, 2 mm[90%, with a slightly better agreement between MC results and measured data compared than Pinnacle Vs Measured data. For the Lung areas, there was a larger discrepancy with MC Vs Measured, with most arcs passing at 3%, 2 mm [90%. Conclusion A MC system has been implemented to allow direct comparison with measured PSQA data and TPS calculations. A good agreement was achieved across all systems, with slightly larger discrepancies evident with lung type arcs.

Introduction In photon radiation therapy, computed tomography (CT) numbers are converted into mass densities (MDs) or relative electron densities to water (REDs) for dose calculation. In order to validate CT-MD and CT-RED calibration tables registered in the radiotherapy treatment planning systems, a postal audit phantom and a stoichiometric CT calibration with three-parameter fit model have [1] . The purpose of our study is to demonstrate an outline of the CT number calibration audit through a multi-institutional study. Method A postal audit phantom was developed with tough lung and tough bone inserts (Kyoto Kagaku, Kyoto, Japan) (Fig. 1) . And, a new stoichiometric CT calibration with empirical three-parameter fit was established. As a pilot study for CT number calibration audit, the postal audit phantom was scanned by six CT scanners at five radiotherapy institutions. The scan conditions were the same as those of the treatment planning CT scan. The obtained CT-MD and CT-RED calibration tables were compared to the CT number calibration tables registered in a radiation treatment planning system (RTPS) at each radiotherapy institutions. The MD or RED differences were compared to the tolerance level, which caused a 2% dose difference. Results The MD or RED differences for each tissue type are shown in Fig. 2 (a)-(c) as stacked histograms. The tolerance levels of lung, adipose/muscle and cartilage/spongy-bone were 0±0.044, 0±0.022 and 0±0.044, respectively [2] [3] . A part of the MD or RED difference for adipose/muscle and cartilage/spongy-bone were exceeded the tolerance levels. However, the mean values of the MD or RED difference were not exceeded the tolerance levels.

Introduction In stereotactic body radiosurgery it is imperative that the isocentre of the linear accelerator is well characterised. The isocentre should be the intersection between the mechanical axis of the gantry, treatment couch and collimator with the imaging isocentre and treatment beam axis [1] . This can be achieved with a Winston Lutz test, wherein a phantom containing a small radiopaque marker is imaged at a series of gantry, couch and collimator positions. Method A Winston Lutz phantom was created in house at the Royal Brisbane and Women's Hospital on Raise 3D Pro 2 dual extrusion printer, using both standard 1.75 mm PLA filament and 1.75 mm Copperfill plastic filament to achieve contrast between ball bearing and outer cube [2] . A Copperfill sphere was printed inside a PLA cube with indents for alignment and optimised ooze conditions to reduce artefacts. Isocentre analysis was achievable using 8 LINAC orientations; 4 Gantry Angles at 270, 0, 90 and 180, collimator angles at 270 and 90 degrees, and two couch angles at 270 degrees and 90 degrees [3] . The phantom was mounted on an xyz stage and fixed in place with an indexing bar. PyLinac Winston Lutz algorithm was used to analyse the images taken at each orientation. Results The phantom was successfully created for routine quality assurance locally. The 8 LINAC orientations chosen provided enough information for the PyLinac Winston Lutz algorithm to produce information about the isocentre of the LINAC. Tracking progress with QAtrack?, over a period of six months, the Gantry 3D isocentre varied by less than 0. 4 mm and the Collimator 2D isocentre and Couch 2D isocentre varied by less than 0.11 mm and 0. 3 mm respectively (Figs. 81, 82) .

Introduction The Fraction-0 module of PerFraction TM (PF) from Sun Nuclear Corporation (SNC) is used as pre-treatment quality assurance(QA) tool in our practice. PF automatically compares the in-air plan delivery of calibrated electronic portal imaging device (EPID) with the predicted plan dose. The calibration method to convert the images to absolute dose is from SNC and establishes EPID response to dose correction factor tables for each: pixel position, field width and primary to scatter contribution along with reference dose information. For PF to perform, monitoring EPID performance is crucial as issues with EPID in addition to the plan delivery systems can result in QA failure. The aim of this study was to follow-up few PF QA failures and identify the necessity for alteration of test frequencies, based on clinic specific experience. Method Different instances of PerFraction IMRT and VMAT QA fails on Varian 21iX and TrueBeam (TB) were investigated, causes for the failures identified and preventive actions discussed. Results Though both 21iX and TB has aS1000 panels, 21iX panel nearing end of its life, indicated frequent failures. Failed cases on this 21iX passed a second exposure on another beam matched linac indicating issues with EPID performance. Since output drifts can change flood field (FF) corrections for an aging panel, a repetition of the dark(DF) and FF resulted in QA pass. This test frequency is now changed to biweekly from monthly. A different set of QA failures revealed that the FF for the linac panel was delivered at a non-zero collimator angle. Repeat delivery of DF and FF at correct collimator angle resolved this. A set of TrueBeam QA failures was resolved by repeating an isocal verification and subsequent calibration. Routine monthly checks being within tolerance, the only parameter to be verified was imager centre drift. Another group of failures on True-Beam required EPID absolute dose calibration. Panel service/ mechanical recalibrations, or output drifts can lead to a change in the SNC stored correction factors. The calibration frequency is altered to biannual from annual on this linac. Conclusion PerFraction TM , a simplified solution for patient specific QA can become time consuming process for Medical physicist facing unexplained QA failures, from unexpected changes in EPID performance. A streamlined QA programme based on TG142 combining TG 100 guidelines to address factors impairing performance quality with increased test frequency can lead to a smooth patient specific QA solution.

Introduction PerFRACTION Fraction 0 TM collects the patient plan data delivered to the EPID, allowing dose reconstruction to be compared via gamma analysis against the TPS dosimetry, prior to the patient treatment. PerFRACTION measurements were performed on three licenced Elekta Agility machines, PA2/PA5/PA6, to investigate panel response and PerFRACTION model accuracy. Method Elekta linear accelerators have a fixed Source-to-Image Distance (SID) of 160 cm with 1024 x 1024 pixels split across 16 subsections, covering the panel size of 41 cm x 41 cm, giving a resolution of 0.4 mm at the SID. Elekta multi-gain panel calibrations for 6X and 10X were performed on all panels prior to the Fraction 0 calibration process, which converts pixel intensity to dose. Gamma criterion of 2%/2 mm was used for checking Fraction 0 calibration and gamma criterion of 3%/2 mm was used for IMRT/VMAT QA analysis. Results The 6X multi-gain panel calibrations were repeated several times on PA6 due to poor gamma pass rates (less than 94%) of the Fraction 0 calibration. In contrast, the gamma pass rate of Fraction 0 Phys Eng Sci Med calibration on PA2/ PA5 were all above 98% for both 69 and 109. Though the calibration passed, the 10X multi-gain panel calibration was repeated on PA2 due to a failure within panel subsections. 2D array measurement with plastic water on PA6 and PerFRACTION measurement on PA2/PA5 showed good pass rate for those IMRT/ VMAT failed PerFRACTION measurement on PA6, indicating the failure lay solely in the PA6 panel, and was not due to a linear accelerator delivery error. Conclusion Baseline reference fields and corresponding gamma pass rate are useful to check the performance of the EPID panels. Per-FRACTION commissioning across three Elekta Agility linear accelerators found the panel response can differ significantly. Per-FRACTION 2D analysis of Fraction 0 TM has been commissioned on PA2/PA5 for clinical use. Introduction SunCHECK TM from Sun Nuclear Corporation is a software package offering tools for EPID and/or log file-based patient-specific quality assurance. SunCHECK's DoseCHECK TM compares dose distribution calculated by TPS against contour-based 3D dose calculation algorithm and can be used as preliminary checks before QA session, indicating eventual plan delivery issues. SunCHECK's Fraction 0 TM reconstructs dose based on log files and/ or EPID image obtained from Linac and compares it against the planning system. The aim of this study is to establish the dose agreement of DoseCHECK and Fraction 0 against the TPS. Method Treatment plan datasets of 156 VMAT cases were exported to SunCHECK server. EPID images were acquired on three treatment machines. 3D gamma was evaluated using 3%/2 mm gamma criterion in DoseCHECK and log-based Fraction 0. Correlation was established using two-tailed Wilcoxon signed-rank test with significance level of 0.05. 3D gamma threshold of 95% was examined against treatment machines and treatment sites (abdomen, breast, lung, chest wall, extremities, head & neck, pelvis & spine). Results 3D gamma pass rate was more than 95% for all machines and treatment sites. Median 3D gamma values for DoseCHECK TM and Fraction 0 was 99.11% and 99.16%, respectively. Four of the studied patients had bolus not included in the external contour resulted in failures but upon contouring the bolus, all four cases passed the set gamma criteria. Conclusion Overall the 3D gamma rate was consistent between DoseCHECK and Fraction 0. Despite a slightly low median gamma values reported by DoseCHECK, it proved to a very good predictor of dose calculation with log-based Fraction 0. As SunCHECK calculation are based on contours as opposed to Pinnacle, any immobilisation devices or scanned in bolus are identified as potential sources of error if not contoured prior to exporting the TPS datasets to SunCHECK.

Introduction The introduction of IMRT and VMAT have promoted performing Patient-Specific QA (PSQA) for each patient before treatment. Several published papers and guideline reviews have addressed the issues of PSQA as well as highlighted its importance. We performed a retrospective analysis of our six years' experience with IMRT/VMAT QA using the ArcCHECK (Sun Nuclear, Melbourne, FL) diode array dosimetry system and reported our departmental guidelines and action levels for IMRT/VMAT delivery. Methods Between March 2014 and March 2020, 318 patients with prostate, head and neck (H&N), brain, pelvis, and other cancer sites were treated with IMRT/VMAT technique at our institution. Following our departmental protocol, for each patient, PSQA is performed prior to treatment. On Varian Eclipse TPS, all clinical plans are recalculated on the ArcCHECK phantom for verification. Following delivery, measured doses are compared to the calculated doses exported from TPS using the SNC patient Software as well as measured doses using a 0.016 cm 3 a 3D pinpoint ionization chamber (PTW-Freiburg). Agreements between measured and calculated doses were evaluated using 3% (global)/3 mm and 95% gamma pass (GP) rate. The mean GP and standard deviations for different sites were evaluated to develop local confidence and tolerance limits. Results For gamma analysis, the average GP rates were 99.3±1.0% for prostate, 98.3±2% for H&N and brain, 98.9±1.2% for pelvis. The overall average GP rate was 98.7±1.5% and the local confidence limit was determined to be 4.0%.The mean point dose differences and standard deviations were 0.4±1.0% for prostate, -0.4±0.9% for brain, 1.2±1.0% for pelvis, and 1.0±1.2% for H&N & other sites. Almost all the point dose results were within our 3% tolerance limit, except for a few variances. Conclusion The study showed that the ArcCHECK dosimetry system permits accurate IMRT/VMAT PSQA. Overall our results indicated that our local confidence limits were in-line with the recommended published literature 1,2 .

factor for the SRS Diode [1] . In addition, the microdiamond OF determination did not require the daisy chaining method which potentially adds additional error, and the good agreement between the ACDS OF and the microdiamond OF increased confidence in this measurement. The microdiamond detector exhibited slightly increased volume averaging thus the SRS diode TMR and OAR data was retained in the TPS. Conclusion The main errors encountered concerned the resolution of the Eclipse dose calculation algorithm and FilmQA Pro analysis software. Measurements and analyses were completed successfully for the commissioning of the 5 mm cone.

Introduction As opposed to CT, MRI offers a superior soft tissue contrast, which is favourable for organ or target delineation in radiotherapy. To obtain the benefits of MRI, while eliminating the planning CT and removing the *2 mm uncertainty introduced by CT-MRI registration, an MRI-only workflow was developed. Already a reality in photon therapy, and promising to one day be used in combination with image guided radiotherapy, the MR-only workflow replaces a conventional CT with a synthetic-CT (s-CT) for treatment planning. Proton therapy, however, is more sensitive to uncertainties associated with s-CTs. Of the many applications of an MRI-only proton treatment planning, we focused on prostate cancer, where lateral opposed beams at 90°and 270°are currently used; however, an alternate beam arrangement using anterior oblique beams has demonstrated benefits. Furthermore, this study utilities a multi-atlas method for generating an s-CT, developed by Dowling et al. [1] , to compare the lateral opposed and anterior oblique beam arrangements. Method This study uses patient data and s-CT generation method from previous work by Dowling et al. [1] . Each patient had a CT and s-CT that was used for treatment planning in matRad [2] . Two treatments plans per patient were generated: lateral opposed fields at 90 and 270 degrees and anterior oblique fields at 30 and 330 degrees. Methods of evaluation were DVH, the dose MAE within main contours, and the range shift of the ten central slices.

Results The dose MAE in the PTV was 0.0185 and 0.0201 Gy for anterior oblique and lateral opposed, respectively. All DVH plots were clinically compliant. Less range shift was recorded for the anterior oblique beam arrangement.

Conclusion This study examined differences between the lateral opposed and anterior oblique beam arrangements and demonstrates that an MRI-only workflow for proton prostate treatment is feasible using a multi-atlas method of s-CT generation.

Purpose The objective of this work was to investigate the absorbed dose in a breast tumour and in neighbouring normal tissue of a mouse by means of a water-equivalent phantom and using Monte Carlo (MC) simulations. Method The mouse phantom was made of Polymethyl Methacrylate (PMMA) to mimic the characteristics of tissue and approximated tumour shape [See Image 1]. The phantom was placed in a specially designed collimator [See Image 2] designed to enable a targeted irradiation using a Cs-137 gamma irradiator [1] . The phantom was irradiated targeting the tumour with doses of 5 to 20 Gy. EBT-XD GAFchromic films were used to measure the dose distribution within the phantom. Additionally, the DOSXYZnrc code was used to compute the 3D dose distribution in the mouse phantom for comparison with the film measurements. Results The dose to normal tissue measured with the film detectors was between 8 to 9% of the tumour dose. Film measurements agreed with the Monte-Carlo calculations to within 2%. Conclusion This physical phantom is a useful tool to examine individual tumour dose and normal tissue dose of a mouse. Introduction With the increase in breath hold type treatments for Lung SABR and Breast treatments, changes in the lung density can influence the accuracy of treatment planning systems. With deep inhalation, coupled with diseased lung, lung densities can drop considerably lower than 0.3 g/cm 3 of ''normal'' lung. This project aims to investigate the accuracy of the Pinnacle TPS for treatments with lowdensity lung. Method In lieu of commercially available solutions for low-density lung analogue, 3D printed slabs of PLA with a Gyroid infill pattern of densities ranging from 5% to 30% were fabricated in-house. Measurements at depths below lung were performed with an ion chamber and compared with Pinnacle and Monte Carlo simulations using BeamNRC/DOSXYZ [1] and MCTools [2] . Results Firstly, validation was performed by comparing the 3D printed gyroid at 30% infill to the commercial lung phantom (Gammex, Middleton WI) and results showed agreement. Measurements and Monte Carlo simulations showed good agreement demonstrating the feasibility of using 3D printed gyroid infill slabs as lung analogues for low lung densities.

There was underestimation of dose under lung in Pinnacle, which varied with density and depth beyond the lung. For example, for the 3 cm thick, 10% infill, a difference of 4.4% to 4.7% was observed at 1-2 mm from the lung interface and this difference reduced to *2% at 2 cm. Dose in lung, using DVH analysis, also showed a detectable difference between Monte Carlo and Pinnacle. Conclusion This investigation has shown that 3D printed lung slabs using a gyroid pattern with variable infill can act as a novel lung equivalent substitute for low lung densities. Comparison with Pinnacle, Monte Carlo and measurements has shown a detectable divergence between Pinnacle calculations and the physical transmission. Further work investigating dose to tumour in lung will also be performed in the future (Fig. 86) .

Introduction Electron treatment in radiation therapy can necessitate patient setup position such that the non-target tissue is in close proximity to the electron applicator side walls. The IEC [1] specifications state that the applicator side plane leakage should not exceed 10% of the central axis dose at D max . Measurements were made to determine the amount of leakage through the sides of the applicator and quantify the thickness of lead for effective shielding to protect the lens or skin. Method A film wrap was initially performed with XRQA-2 on a 10 9 10 cm 2 applicator on a Varian Clinac iX to identify hotspots. Measurements using various dosimeters were performed for 6 -18 MeV at 5 -8 cm from the projection of field edge. The ionisation ratio and dose were measured using an Advanced Markus chamber and PTW 60017 Diode E respectively and in water equivalent material. Ratios are normalised to D max at Gantry 0°and 100 cm SSD (Fig. 87) . Results Leakage dosimetry was quantified with an ionisation ratio curve measured using an Advanced Markus for 0 -10 mm and it was comparable to the dose readings acquired using the PTW Diode E. The maximum peripheral leakage from applicator side walls was \5% of the central-axis dose maximum, with the highest dose leakage of 3.1% at 5 ± 0.5 mm. The ionization ratio and diode dose ratio at 10 mm depth were within 0.5%. Adding a wax coated piece of 10 kg/m 2 lead lowered the leakage to less than 1% for all energies. Conclusion The leakage doses from the sides of a Varian iX electron applicator are low and can be effectively reduced further with 10 kg/ m 2 lead eye or skin shielding. Method A Varian clinac was previously modelled and validated 1 for 6 MV and 10 MV beams using the EGSnrc/BEAMnrc Monte Carlo (MC) code 2 . The same MC input files were used for FF beams, while the flattening filter was removed for FFF beams. For each energy, particle information from MC simulations of a 10 cm 9 10 cm field were stored in a phase space file and the fluence, energy spectra and electron contamination were evaluated. Phase space files were then used to calculate PDD in water using the DOSXYZnrc MC code. Surface dose in the build-up region were evaluated for all FF and FFF beams. Results Significant differences in the energy fluence were observed between FF and FFF beams for both 6 MV and 10 MV (Figure 1 ). From the phase space files, the amount of electron contamination was 0.30% for 6 MV-FFF, 0.35% for 6 MV-FF, 0.40% for 10 MV-FFF and 0.50% for 10 MV-FF. PDD calculations showed 15% higher surface dose for 6 MV-FFF compared to 6 MV-FF and 25% higher surface dose for 10 MV-FFF compared to 10 MV-FF. Conclusion Significant differences in energy fluence were observed between FF and FFF beams. Electron contamination for FF and FFF beams are similar. The higher surface dose for FFF beams is therefore due to its softer spectrum compared to FF beams.

Introduction CT chest abdomen pelvis (CAP) studies with contrast deliver some of the highest radiation doses in diagnostic CT. Thus optimising the scan parameters to lower the radiation burden on the Figure 1 Energy fluence distributions for 10 MV-FF, 10 MV-FFF, 6 MV-FF and 6 MV-FFF beams Phys Eng Sci Med patient becomes paramount for reducing associated radiation risks. One method of lowering the radiation dose in CT CAP is to conduct the entire scan in a single-phase acquisition, rather than a chest scan followed by an abdomen pelvis scan. The single phase approach reduces the amount of tissue that is scanned in the overlap between the two acquisitions. This study aims to compare the organ dose and effective dose to patients undergoing CT CAP studies using a dualphase approach versus a single-phase approach. Method A retrospective audit of trauma patients undergoing singlephase CAP and dual-phase CAP on the same CT scanner was conducted. The radiation dose was simulated on CT Expo v2.5, and a student T-test was conducted to assess the significance between both groups. Results The radiation dose was significantly lower for both male and female patients undergoing a single-phase acquisition. Additionally, the radiation dose to the breast tissue was significantly decreased using the single-phase approach.

Conclusion Radiation dose optimisation in trauma patients requiring CT CAP with contrast is feasible by utilising a single-phase scanning approach. The dose reduction due to not scanning the same organs twice due to overlap outweighs the absorbed dose due to the slightly increased radiation output through the chest region.

Introduction Local diagnostic reference levels (LDRLs) are an important tool to support the optimisation of imaging examinations. There are currently no published Australian Diagnostic Reference Levels for general x-ray examinations for paediatric patients. Local DRLs have been established for the most frequent and relatively higher dose examinations and compared to published international DRLs [1] for the Queensland Children's Hospital (QCH). Method A total of 110,000 patient exposure records were included in this assessment. The LDRLs have been defined at the 75 th percentile for five units in the department grouped together [2] . This paper presents the LDRL values for general radiography for 15 examinations, where there was adequate data, across six age groups, given in Table 1 . The anatomical regions for which adequate data was available are head, chest, abdomen, pelvis, cervical spine, thoracic spine, lumbar spine and whole spine, with AP/PA and lateral projection data. The dosimetric quantity used to set the LDRLs was the dose area product (DAP) in units of mGycm 2 .

Results The results of the LDRL analysis are given in Table 2 : Table 2 Results for all projections for LDRL data These results have been summarised and compared to published international DRLs. The LDRLs were benchmarked against the DRLS defined by the European paediatric DRLs [1] (EDRLs) where available. The LDRLs for paediatric general x-ray examinations are lower than the published EDRLs for all age groups and available examinations. Conclusion Establishing LDRLs for general x-ray procedures is useful for providing a baseline for future optimisation of exposure levels. It has shown that QCH has LDRLs that are lower than published European data. Introduction Clinical Governance processes informed by Statistical Process Control (SPC) tools have been shown to promptly flag potentially significant variations in clinical performance in a range of health care settings. Such systems usually rely on data from dedicated clinical registries, however, these are traditionally expensive to establish and maintain. Data from administrative systems is viewed as a viable alternative data source. We discuss the application of these techniques to inform the morbidity and mortality (M&M) processes in Orthopaedic Surgery at a single site.

Method Routinely collected coding, financial and limited adverse outcome screening (LAOS) data for acute orthopaedic surgical admissions, between January 2013 and the March 2020, were retrospectively evaluated. A range of process and outcome performance from coding must be used with caution due to the nature of the data source. Use of monitoring tools based around these measures may best be used as a screening tool for identification of cases for detailed review.

Conclusions SPC tools facilitate near ''real-time'' performance monitoring allowing early detection and intervention in instances of altered performance. While having some limitations by comparison to analysis using prospectively collected clinical registry data, careful and considered review of analysis may prove helpful in detecting and differentiating systemic and individual clinician variation.

Phys Eng Sci Med

Process-based quality management for clinical implementation of adaptive radiotherapy

Adaptive Radiotherapy: Moving Into the Future

Patient-specific validation of deformable image registration in radiation therapy: Overview and caveats

Use of image registration and fusion algorithms and techniques in radiotherapy: Report of the AAPM Radiation Therapy Committee Task Group No. 132

AtlasNet: multi-atlas non-linear deep networks for medical image segmentation

Institutional Patient-specific IMRT QA Does Not Predict Unacceptable Plan Delivery IJROBP

Development and testing of an improved dosimetry system using a backscatter shielded electronic portal imaging device

P Ramachandran Radiation Oncology, Princess Alexandra Hospital, Ipswich Road

Investigation of a real-time EPID-based patient dose monitoring safety system using site-specific control limits

Patient-specific online dose verification based on transmission detector measurements

Quality assurance for nonradiographic radiotherapy localization and positioning systems: Report of Task Group 147

Surface Guided Radiation Therapy

Assessment of the accuracy of truebeam intrafraction motion review (IMR) system for prostate treatment guidance

483 -Dosimetry of Small Static Fields Used in External Beam Radiotherapy An International Code of Practice for Reference and Relative Dose Determination

An analytical formalism for the assessment of dose uncertainties due to positioning uncertainties

A 2D silicon detector array for quality assurance in small field dosimetry: DUO

Characterization of ELEKTA SRS cone collimator using high spatial resolution monolithic silicon detector array

Shading correction for on-board cone-beam CT in radiation therapy using planning MDCT images

The reconstruction toolkit (RTK), an open source cone-beam CT reconstruction toolkit based on the insight toolkit (ITK)

ScatterNet: A convolutional neutal network for cone-beam CT intensity correction

Absorbed dose calorimetry

Electron contamination modeling and skin dose in 6 MV longitudinal field MRIgRT: Impact of the MRI and MRI fringe field

Dosimetric Optimization and Commissioning of a High Field Inline MRI-Linac

Technical Note: Experimental results from a prototype high-field inline MRI-linac 278

Accuracy, repeatability, and interplatform reproducibility of T1 quantification methods used for DCE-MRI: Results from a multicenter phantom study

Quantitative Imaging Biomarkers Alliance (QIBA) recommendation for improved precision of DWI and DCE-MRI derived biomarkers in multicenter oncology trials

ISMRM/NIST MRI System Phantom. Collaborate NIST

Assessing effects of scanner upgrades for clinical studies

O070 Out of field dose due to electrons generated infield on an Elekta Unity MR linac J Baines

Use of image registration and fusion algorithms and techniques in radiotherapy: Report of the AAPM Radiation Therapy Committee Task Group No. 132. Medical physics

A model to accumulate fractionated dose in a deforming organ

Is it sensible to ''deform'' dose? 3D experimental validation of dose-warping. Medical physics

A novel methodology for 3D deformable dosimetry. Medical physics

Nanomechanical analysis of cells from cancer patients

Accuracy of finite element model-based multiorgan deformable image registration. Medical physics

Biomechanics: mechanical properties of living tissues

Advances in kilovoltage x-ray beam dosimetry

Data for the dosimetry of low-and mediumenergy kV x rays

SpekCalc: a program to calculate photon spectra from tungsten anode x-ray tubes

O083 A method of assessing and correcting for temporal changes in optical density using radiochromic film: Can we get instant results?

Centre for Radiation Oncology

Investigation of lung tumour peripheral doses using normoxic polymer gel and film dosimetry techniques

Optimum photon beam procurement in modern radiotherapy departments 2019 ESTRO meets Asia

Investigation of the PAGAT polymer gel dosimeter using magnetic resonance imaging

Optimization of multiple spin-echo sequences for 3D polymer gel dosimetry

Prabhakar Ramachandran Radiation Oncology, Princess Alexandra Hospital

The Use of Alanine as a Solid Dosimeter The Use of Alanine as a Solid Dosimeter

Dosimetry by ESR spectroscopy of alanine

Alanine dosimetry at NPLthe development of a mailed reference dosimetry service at radiotherapy dose levels

Absorbed-dose/dose-rate dependence studies for the alanine-EPR dosimetry system

O089 Plasma therapy for cancer: a new approach for delivery of active agents without radiation beams

Synergy between non-thermal plasma with radiation therapy and olaparib in a panel of breast cancer cell lines

2 School of Pharmacy

Stably engineered nanobubbles and ultrasound-An effective platform for enhanced macromolecular delivery to representative cells of the retina

Concentration stability of monolayer and bilayer lipid microbubbles-a precursor for new drug delivery applications

Ultrasoundresponsive nanobubbles for enhanced intravitreal drug migration: An ex vivo evaluation

Attenuation and size distribution measurements of Definity TM and manipulated Definity TM populations

Method for Preparing a Lipid Bubble

The next step in personalised medicine?

Potential Gender Differences in a Normal Tissue Complication Probability Model for Heart Toxicity During Radiation Therapy for Esophageal Cancer

3 Hunter Medical Research Institute (HMRI) Imaging Centre

The Physics of Radiation Therapy. 2 nd Edition

State of the art on dose prescription, reporting and recording in intensity-modulated radiation therapy

Fundamental quantities and units for ionizing radiation

) documents including: Monaco Dose Calculation Technical Reference and letter titled ''SUMMARY OF DOSE TO WATER CALCULATIONS IN MONACO (V3.1 and later)'' dist

Dose specification for radiation therapy: dose to water or dose to medium?

Practical IMRT QA dosimetry using Gafchromic film: a quick start guide. Australasian physical & engineering sciences in medicine

The Physics of Radiation Therapy. 2 nd Edition

State of the art on dose prescription, reporting and recording in intensity-modulated radiation therapy

Fundamental quantities and units for ionizing radiation

) documents including: Monaco Dose Calculation Technical Reference and letter titled ''SUMMARY OF DOSE TO WATER CALCULATIONS IN MONACO (V3.1 and later)'' dist

Dose specification for radiation therapy: dose to water or dose to medium?

P014 Combined electron paramagnetic resonance (EPR) and optically stimulated luminescence (OSL) investigation of the defect states in BeO dosimeter chips Christopher Noble, Prabhakar Ramachandran Radiation Oncology

Review of optimal radiotherapy utilization rates. Ingham Institute for applied medical research

Radiotherapy: The Tyranny of distance

Techniques and Technologies in Radiation Oncology

Radiotherapy Treatment Services to NSW Residents

quality-improvement-tools P016 Characteristics of a single CT-ED Curve for all Varian Halcyon linear accelerators

4 Director of Therapeutic Physics, Princess Alexandra Hospital

Advances in radiation therapy

Deep Learning for Medical Image Processing: Overview, Challenges and the Future

Deep Learning Techniques for Medical Image Segmentation: Achievements and Challenges

Cham P018 Can we reduce imaging time and still generate acceptable Substitute CT for Prostate MRI Only Treatment Planning?

Presenting author]). 2 CSIRO Health and Biosecurity, The Australian e-Health & Research Centre

Dedicated Magnetic Resonance Imaging in the Radiotherapy Clinic

Use and uncertainties of mutual information for computed tomography/magnetic resonance (CT/MR) registration post permanent implant of the prostate

MRI-guided prostate radiation therapy planning: Investigation of dosimetric accuracy of MRI-based dose planning

Automatic substitute computed tomography generation and contouring for magnetic resonance imaging (MRI)-Alone external beam radiation therapy from standard MRI sequences

P019 Implementation of quality assurance procedure for deep inspiration breath hold (DIBH) treatments

Gibbs Radiation Oncology Princess Alexandra Raymond Terrace (ROPART)

Commissioning and Routine Quality Assurance of the Vision RT AlignRTÒ System. In: Hoisak J (ed) Surface Guided Radiation Therapy

Vision RT Calibration Phantom User Guide

P021 Review of ion chamber array for photon beam quality index and electron energy measurements V Seshadri, S Ibrahim, P Ramachandran Radiation Oncology

P024 The use of a Monte Carlo system for regular comparison with patient specific QA measurements and TPS calculations for SBRT treatments

Presenting author]). 2 RBWH -Royal Brisbane and Women's Hospital

BEAM: A Monte Carlo code to simulate radiotherapy treatment units

GLaS: An absolute dose calibration algorithm for an amorphous silicon portal imager. Applications to IMRT verifications

P025 Development of a CT number calibration audit phantom and the tolerance levels for each tissue type

Hideharu Miura 1,2 , Kiyoshi Yamada 1 , Kosaku Habara 1 , Fumio Okamura 1 , Daisuke Kawahara 1

Presenting author]). (hayata@hiprac.jp), (ozawa@hiprac.jp), (miura@hiprac.jp), (yamada@hiprac.jp), (habara@hiprac.jp), (okamura@hiprac.jp). 2 Hiroshima University

jp), (rt.ymorimoto@gmail.com). 5 Hiroshima City Hiroshima Citizens Hospital, Hiroshima, Japan. (kawakubo6405211@gmail.com). 6 Hiroshima Red Cross Hospital & Atomic-bomb Survivors Hospital, Hiroshima, Japan. (hnozaki@hiroshima-med.jrc.or.jp). 7 Hiroshima Prefecture Cancer Control Division

Development of a CT number calibration audit phantom in photon radiation therapy: A pilot study

Tolerance levels of CT number to electron density table for photon beam in radiotherapy treatment planning system

Tolerance levels of mass density for CT number calibration in photon radiation therapy

P026 3D printed Winston Lutz phantom for LINAC quality assurance

Crowe 1,2 1 Cancer Care Services, Royal Brisbane and Women's Hospital, Brisbane. (Emily.simpson-page@health

Quantifying isocenter measurements to establish clinically meaningful thresholds

Radiological properties of 3D printed materials in kilovoltage and megavoltage photon beams

On the selection of gantry and collimator angles for isocenter localization using Winston-Lutz tests

Jaysree Ukath 1 , Penny Fogg 1 , Brindha Subramanian 1

Ukath@genesiscare.com), (Penny.Fogg@genesiscare.com), (Brindha.Subramanian@genesiscare.com). 2 Director

Task Group 142 report: Quality assurance of medical accelerators

The Report of Task Group 100 of the AAPM: Application of Risk Analysis Methods to Radiation Therapy Quality Management

P Ramachandran Radiation Oncology, Princess Alexandra Hospital, Ipswich Road

Tolerance limits and methodologies for IMRT measurement-based verification QA: Recommendations of AAPM Task Group No. 218

P030 Patient-specific QA for IMRT/VMAT treatment plans: a retrospective analysis of six years' experience Y Yousif

IMRT commissioning: Multiple institution planning and dosimetry comparisons, a report from AAPM Task Group 119

Guidelines for the verification of IMRT

TRS-483 Dosimetry of Small Static Fields Used in External Bram Radiotherapy: An International Code of Practice for Reference and Relative Dose Determination

MRI)-only prostate proton treatment planning: comparison of lateral opposed and anterior oblique beam arrangements S Hickey 1 , A L Fielding 1 , J A Dowling 2,3 , P B Greer 3,4 , J Seco 5,6 1 Faculty of Science and Technology

4 Calvary Mater Newcastle Hospital

'matRad, e0404, Division of Medical Physics in Radiation Oncology

OpenMP: an industry standard API for shared-memory programming

Fast calculation of the exact radiological path for a three-dimensional CT array

Modelling the effect of geometric uncertainties, clonogen distribution and IMRT interplay effect on tumour control probability (Doctoral dissertation

Institute of Health & Biomedical Innovation

A novel add-on collimator for preclinical radiotherapy applications using a standard cell irradiator: design, construction, and validation

3D Slicer: a platform for subject-specific image analysis, visualization, and clinical support

BEAM: A Monte Carlo code to simulate radiotherapy treatment units

P039 Can 3D printed phantoms accurately model lowdensity lung? Investigating the accuracy of Pinnacle

Sylwia Zawlodzka-Bednarz ROPART -Radiation Oncology Princess Alexandra Hospital Raymond Terrace

BEAM: A Monte Carlo code to simulate radiotherapy treatment units

P040 Assessment of out of patient plane leakage doses in high energy electron beams for a Varian Clinac iX J Miller, L Sim Radiation Oncology Princess Alexandra Raymond Terrace, Princess Alexandra Hospital

Evaluation of Clinical Dose Distributions Using Monte Carlo Methods

BEAM: A Monte Carlo code to simulate radiotherapy treatment units

Yohei Inaba 1,2 , Shin Hitachi 3

). (chida@med.tohoku.ac.jp) Introduction The physicians are close to the source of scattered radiation dose during CT-guided interventions. Therefore, dose-reducing/optimizing for physicians are important. Furthermore, the occupational dose limit to the eye lens is reduced approximately 1/8. The physician and medical staff are at high risk of radiation-induced injury, such as cataracts. However, few previous reports have investigated radiation doses to physicians in MDCT-guided interventions. The purposes of this study were to evaluate the radiation dose to the physician and medical staff during CT-guided interventions. Method The procedures were performed using IVR-CT system. We evaluated the occupational radiation dose (physician and medical staff) and related parameters (fluoroscopy time, mAs, CTDI and DLP) performed consecutive about 200 procedures in CT-guided biopsy. Results Physician doses (eye lens, neck and hands; lSv, average ± SD) in our MDCT-guided interventions were 39.1 ± 36.3, 23.1 ± 23.7, and 28.6 ± 31.0, respectively. Medical staff (nurse) doses were lower than the physician doses, obviously. Acquisitions numbers, fluoroscopy time, fluoroscopy mAs (average ± SD) of our MDCTguided interventions were 34.7 ± 23.1, 26.6 ± 17.8s, and 650.0 ± 598.4, respectively. Conclusion We have understand the importance of reducing/optimizing the dose to the physician and medical staff in CT-guided interventions

Michael.lawson@monashhealth.org [Presenting author]), (Ahilan.Kuganesan@monashhealth.org). 3 Department of Medical Imaging and Radiation Sciences, School of Primary and Allied Health Care, Faculty of Medicine

Best protocol for combined contrast-enhanced thoracic and abdominal CT for lung cancer: A single-institution randomized crossover clinical trial

The establishment and use of dose reference levels in general paediatric radiology

3 Brisbane Hip Clinic

Acknowledgements The help of the TROG QA team as well as the financial support of Breast Cancer Trials and the Gross Foundation is greatly appreciated.O007 A freely available standardized adaptive evaluation report template Acknowledgements I would like to thank the Royal Children's Hospital in Melbourne for providing the phantom used in the study. I would also like to thank the Peter MacCallum Cancer Centre for providing the TLD dosimeters and read-out system. Acknowledgements The authors are grateful to all staff at Calvary Mater Newcastle Hospital who collected these data. References Acknowledgements The authors want to thank iRT for loan of the IQM transmission detector system and their support during installation and initial test of the system. References Acknowledgements We thank the AAPM, Scandidos and Varian for their support.

Calvary Mater Newcastle is a reference site for C-Rad. References Acknowledgements We would like to thank GenesisCare for access to their linacs.

The authors would like to acknowledge the many authors that contributed to the publication, and in turn the many more that contributed towards the published data. The authors of the paper not included as authors on the abstract are: D 

signal for both 6X and 10X (d max 1.5 and 2.2 cm) and avoid the buildup region. Results One example of EPID vs Pinnacle for a modulated field is shown in Figure 1 . Gamma 3%/3 mm resulted in 76.83% of points \ 1. The profile comparison shows differences in absolute dose. The open field at same angle resulted in 79.44% gamma points \ 1. Using Octavius,100.0% of gamma points were \ 1 ( Figure 2 ) with same criteria for all the open and modulated beams. Conclusion The EPID/EPIQA is not accurate to provide pass rates greater than 90%, whereas the results for the Octavius system demonstrate it is a reliable tool for DIBH QA. The inaccuracies of the EPID system will be discussed. References/Acknowledgements 1. Nicolini, G., Fogliata, A., Vanetti, E., Clivio, A. and Cozzi, L.( Introduction Gamma Knife Icon provides stereotactic radiosurgery treatments. Patients are fixated with a thermoplastic mask, and motion during treatment is monitored via its infra-red high definition motion management (HDMM) of a marker placed on the patient's nose. The aim of this study was to evaluate the movement of patients during treatment, and to correlate this movement to the displacement of the target volumes during treatment.Method Three patients presented with multiple metastases were chosen as an initial dataset. Target displacement was analysed via log-files generated by the HDMM system. The patient movement is transformed to the movement of the target volumes by applying matrix transformation (rotations and translations) to the target volumes, exported in the DICOM format, corresponding to the transformation from reference point to deviation point. The rotation pivot point was identified as a fixed point at the back of the skull by a radiation oncologist. Furthermore, the direction of movement was also evaluated and compared between patients to evaluate a possible trend in the shift.Results Initial results indicate that the modelling of the patient movement is highly dependent on several factors, particularly location of the pivot point and assumptions regarding the relative contribution of rotation and translation to target displacement. The measured target motion varied for the three patients evaluated.Conclusion Several factors impacted the results and they should be further evaluated -mainly the selection of the pivot point and the relative contirubtions of rotations and translations applied. However, the initial results show that a substantial relative displacement could occur in fractionated stereotactic radiosurgery treatments, especially for small tumour volumes. Time of displacement also needs to be incorporated for reasonable results. Introduction Conical collimators are effective and readily available accessories for the field shaping of small stereotactic fields, however the measurements required to accurately characterise the smallest radiation fields are difficult and prone to large errors. Furthermore, there is little published commissioning data to compare measurements against. The aim of this investigation was to measure the required Eclipse algorithm data, being tissue maximum ratios (TMR), off axis ratios (OAR) and the output factor (OF) for a 5-millimetre conical collimator attached to a Varian TrueBeam linear accelerator using a microdiamond detector, and compare these to SRS diode measurements, Varian Golden Beam Data (GBD), an ACDS audit measured OF, radiochromic film measurements and the Varian Eclipse treatment planning system (TPS) calculated data.Method A PTW BEAMSCAN water tank phantom was used to collect the PTW60019 microdiamond data, with the output correction factor derived from TRS-483 [1] . These were compared to the PTW60018 SRS diode measurements, GBD and an ACDS OF. Gamma analysis was used to compare the measured TMRs and profiles, as well as the Eclipse dose planes to the scanned films.Results It was found that most of the data agreed well, with the microdiamond OF being preferred over the SRS diode OF for the TPS configuration due to the availability of a published small field correction factor in TRS-483 compared to an unpublished correction Introduction This talk discusses differences in CT to density and Electron density curves used for treatment planning across a range of treatment planning systems and CT scanners. It also highlights the differences encountered nationally between phantom size and manufacturer. Results of similar investigations have been published 1,2 , but have not included a range of CT scanners and phantoms used in different departments. Method CT to Density or electron density curves for a range of treatment planning systems and CT scanners across GenesisCare in Australia were collected. CT scanner response curves were compared across a kV range of 120kV to 135kV with a range of reconstruction algorithms used clinically. Points of difference were identified and investigated.Results Three main areas of difference identified were the choice of manufacturer (CIRS 062M Vs Gammex 467), small vs large body phantom, and number of bits of the CT scanner (12 Vs 16 bit).Conclusion Across a range of CT scanners and phantoms, the main difference in CT to Density or electron density curves is not only limited to the choice of CT scanner or choice of kV above 120 kV, but also to the choice of CT to ED phantom manufacturer. The use of CT scanners in extended range (for 12 and 16 bit scanners) does not accurately reflect the density for materials with HU greater than 2850 HU. Hence materials must be considered metal and overridden with the correct density of the material used.References/Acknowledgements Method HU to q e relationship was obtained from Pinnacle TPS. The CT image was resampled to 3 mm isotropic spacing. A ray was created at source position for given beam and traced through all voxels using classic Siddon's raytracing algorithm [2] . The algorithm treated a 3D CT image as a set of orthogonally placed parallel planes as shown in Figure 1 . A sequential RPL was already available in our in-house developed software [3] . The nested for-loop was parallelized using OpenMP multithreading. The results were tested on virtual and anonymized CT images from different anatomical sites.Results Improvement in calculation speed was approximately 3.5 with a 4-core CPU. Furthermore, we found the accuracy of results is around 1% compared to sequential method which was within clinically acceptable tolerance limits. The differences in calculation speed for sequential and parallel computing methods as a function of number of voxels are shown in Figure 2 . Introduction Current studies investigating the application of radiotherapy treatments of mice use a wide variety of irradiation techniques. Some techniques irradiate a large percentage of the mouse volume, which can influence the accuracy of the study due to biological effects that differ from a more targeted treatment that would exist with human patients. The aim of this study was to work towards a research platform that can use more targeted radiotherapy to better emulate patient type treatments using mouse models. Method A micro CT of a mouse with a flank tumour was segmented and converted to a 3D printable file (STL) using Slicer [1] . Multiple phantoms were created and 3D printed using PLA filament. Both homogenous water equivalent phantoms, and phantoms with a skeleton of higher infill density printed in-place were created with splits in the model to allow film placement in the tumour area. A treatment plan of the beam geometry was generated on custom code and McToolkit [2] for Monte Carlo simulations based on clinical CT of the printed phantoms. The dose distribution from Dosxyz/ BeamNRC [3] using a previously commissioned superficial model of an Xstrahl 150 treatment unit and compared with film measurements.

The process of taking a Micro CT of a mouse with a tumour ? 3D printed phantom ? Monte Carlo simulation, and measurements using film have been demonstrated. Further work to optimise geometry and treatment types depending on the tumour location can now be performed with the goal towards large scale mouse studies using a clinical superficial treatment unit and 3D Monte Carlo TPS for planning using free open source software (Fig. 85) Introduction There were multiple reports on various artifacts found on x-ray images involving neonates in the incubator unit at the Royal Brisbane & Women's Hospital. An investigation was initiated to find the origin of these artifacts. From the screening process, it was discovered that these artifacts to be a result of manufacturing defects within the detector tray and scale component of the incubator unit. Method A systematic approach was used to identify the source of the artifacts on the x-ray images taken of the neonates inside an incubator unit. Firstly, all CR cassettes and other materials commonly used inside the incubator unit, such as blanket and mattresses, were screened to ensure that they are free from artifacts. Secondly, components identified within the x-rays primary beam were dissembled from the main incubator unit and screened to check for foreign materials or defects that may cause artifact appearance. All x-ray images taken from the screening were subjectively analysed as per criteria set by the radiologist.Results From this investigation, it was discovered that artifacts in the x-ray images of the neonates resulted from manufacturing defects within the detector tray and the scale component of the incubator unit. Conclusion Following this investigation, a plan to screen all incubator units in NICU is put in place to identify the extent of manufacturing defects for TGA [1, 2] reporting and isolate any defective components that may cause artifacts in the neonate x-ray images. The fleet screening process will also be performed in stages to minimise the impact on the clinical operation. In this process, all defective components will be returned to the manufacturer, and a replacement will be requested. Medical physics is also encouraged to be involved in the future acceptance process involving medical equipment such as neonates incubator unit before the unit is released into clinical use. This is so the care and safety for each neonate are kept to the highest level by reducing the number of repeat x-rays while maintaining medical diagnosis integrity. Background Regulation 4R ''Importation of Radioactive Substances'' of the Customs (Prohibited Imports) Regulations 1956, as amended, deals with shipments of radioactive material entering Australia. The CEO of ARPANSA or an authorised officer as appointed by the Minister can approve the importation into Australia of a radioactive substance. A radioisotope is considered to be for medical use when it is intended to be:1. administered to humans for diagnostic or therapeutic procedures• used in any in-vitro medical diagnosis or test • used in research which is either directly or indirectly related to medical diagnosis or therapy Sealed and unsealed radioactive sources used to calibrate instruments in medical practices and pathology laboratories are also included as medical radioisotopes for permit purposes. Aims This paper will highlight the methodology that ARPANSA employs to ensure the efficient authorisation of importations of medical radioisotopes into Australia. Methods Under the Customs (Prohibited Imports) Regulations, a permit from ARPANSA is required for the import of radioactive material. Applications for the import of radioisotopes for medical use must be made to ARPANSA on the appropriate form. These applications can be for either a single shipment or, under certain conditions, for an unlimited number of shipments over 12 months. Results ARPANSA processes hundreds of permits a year, predominately for medical therapeutic uses in brachytherapy and radiopharmaceutical therapy with unsealed sources. ARPANSA's ability to be flexible when dealing with clients, hospitals and state radiations regulators ensure uninterrupted access to radioisotopes in Australia.Conclusion The importation of medical radioisotopes is vital for the health and wellbeing of Australian patients. ARPANSA recognises the need for a quick and efficient service that puts the needs of patients at the forefront of the process.