JPGM_March04.pmd 62 J Postgrad Med March 2004 Vol 50 Issue 1� � 62 CMYK Executive Medical Director, Mirada Solu- tions, Oxford, United Kingdom; *Medical Director, Hunter Area Pathology Service, Newcastle, and Professor and Head, Discipline of Anatomical Pathology, University of Newcastle, Australia. Correspondence: Prof Anthony S-Y Leong, MD E-mail: aleong@mail.newcastle.edu.au www.jpgmonline.com Received : 19-09-03 Review completed : 20-10-03 Accepted : 22-10-03 PubMed ID : 15048004 J Postgrad Med 2004;50:62-69 Digital photography in anatomical pathologyDigital photography in anatomical pathologyDigital photography in anatomical pathologyDigital photography in anatomical pathologyDigital photography in anatomical pathology Leong FJW-M, Leong ASY* ABSTRACTABSTRACTABSTRACTABSTRACTABSTRACT Digital imaging has made major inroads into the routine practice of anatomical pathology and replaces photographic prints and Kodachromes for reporting and conference purposes. More advanced systems coupled to computers allow greater versatility and speed of turnaround as well as lower costs of incorporating macroscopic and microscopic pictures into pathology reports and publications. Digital images allow transmission to remote sites via the Internet for consultation, quality assurance and educational purposes, and can be stored on and disseminated by CD-ROM. Total slide digitisation is now a reality and will replace glass slides to a large extent. Three- dimensional images of gross specimens can be assembled and posted on websites for interactive educational programmes. There are also applications in research, allowing more objective and automated quantitation of a variety of morphological and immunohistological parameters. Early reports indicate that medical vision systems are a reality and can provide for automated computer-generated histopathological diagnosis and quality assurance. KEY WORDS: Digital photography, anatomical pathology, Internet, medical vision systems, education, computer- generated diagnosis, pathology reporting, quality assurance, remote consultation mong the many functions of an anatomical pathologist ar e diagnosis, consultation, documentation, and edu- cation. Implicit in these activities is the need to record mor- phological findings both at the macroscopic and microscopic level.1 Conventionally, this has been performed through the process of descriptive prose with inherent variations in the style, vocabulary and abilities of individual pathologists, which are often considerable and beset by idiosyncrasies. Comparisons and descriptions associated with food are commonplace with such confusing classics like “caseous necrosis” which describes a resemblance to cottage cheese and not cheddar cheese, “nut- meg liver” resembling the cut surface of a nutmeg seed and not the fruit, and “anchovy sauce” which may not be found on the dining tables in many countries. Common colours may become irreproducible when terms like “off white”, “light grey ”, “creamy-white”, “pale white”, “grey-tan”, “egg white” and “ice- white” may be used for the same colour. Photographs documented the true appearances of the patho- logical changes and eliminated much of the inaccuracies re- sulting from the variation in descriptive prowess but there re- mained the inevitable delays of print production and cost re- stricted the number of photographs. Mostly black and white prints were employed in reports. Some institutions maintained elaborate photography set-ups that depended on trained staff to process and print these photos, and to develop coloured projection slides. The advent of the Polaroid film allowed in- stant prints and also immediate projection slides averting the problems of delays, but again costs were significant and qual- ity was not optimal. The development of digital photography and the rapidly fall- ing costs of good quality digital cameras has made major in- roads into our traditional way of documenting pathological findings at both the gross and microscopic level. Together with the expanding use of the Internet, digital photography paves the way for many new applications of image documentation in anatomical pathology. This review will be confined to the capturing of digital images and their current and potential applications in anatomical pathology. Advantages of Digital Photography There are several compelling reasons to convert to digital tech- nology but the most persuasive is lower running costs and shorter image production turnaround times. Following the initial capital investment, additional costs ar e minimal. Memory cards for temporary storage of images are reusable and storage media such as CD-ROMs are inexpensive. Results are almost instantaneous and images can be edited, and the quality and composition improved. It is easier to catalogue, archive and disseminate/transmit a digital image than a print or 35-mm slide. Multiple digital copies in different formats are possible. Photographic laboratories can create glossy prints A Education Forum 63J Postgrad Med March 2004 Vol 50 Issue 1 � � C MYK 63 that are virtually indistinguishable from those made from a chemical film. In addition to improving current practice, the change to the digital medium opens up a new range of applications such as telemicroscopy/ telepathology, image processing and analysis, digital slide archiving and, eventually, automated machine vi- sion systems. Another reason is necessity. As other disciplines of medicine and non-medical fields convert to the digital me- dium, lack of familiarity with digital imaging may eventually become a handicap and be viewed as a deficiency in practice. The Digital Camera Many of the factors important in chemical photography are not applicable to digital photography. Consideration of film speed and type are not relevant and factors such as colour, tem- perature and white balancing are considerably easier to com- pensate for with digital imaging. Focussing is much easier as many cameras have automatic fine focussing. Three primary functions of the microscope are to resolve de- tail, increase contrast and magnify. Digital imaging can en- hance the performance of the microscope through its ability to improve upon the latter two functions, the first being still dependent on optics. The variables influencing image quality are too numerous to discuss but good optics remains the sin- gle most important factor in the acquisition of high-quality digital images. Optical deficiencies revealed by chromatic vari- ance and poor image clarity are far more noticeable when us- ing a digital image sensor compared with 35-mm film. There has been a proliferation of digital microscopy cameras in recent years and a short list of the commonly advertised cameras by the larger microscopy companies is provided in Table 1. Although such devices are generally maintenance-free, it is preferable to select a camera from a manufacturer, which reviews and updates its software drivers with the same degree of efficiency and frequency as a major computer peripheral manufacturer. This is particularly necessary given the rapid rate of evolution of computer operating systems. Cameras fall into three main groups ranging from budget to expensive, and those for specific requirements, e.g. low-light imaging, multi-spectral imaging.2 The cameras listed here ar e generally capable of reasonable imaging for transmission, pres- entation and print. The more expensive models mostly pro- duce larger images of higher quality. Currently, the acquired image size is equated with quality but in the future, the cru- cial determinant will be the method by which the image is acquired at the sensor level as this has a greater influence. Not listed are cameras optimised for low-light photography and cameras capable of high-resolution, high frame-rate imaging. The latter have the advantage of using a technique of multi- ple-frame averaging to improve image quality. As such cam- eras have specific applications, additional e xpense can be avoided by ensuring that specialised features are not purchased if not requir ed. High-end general microscopy cameras such as the ProgRes C14 and Nikon DXM1200 are capable of respect- able low-light imaging. There are essentially two groups of digital microscopy cameras – those designed for still image capture and those for ‘live’ image display. The latter gained popularity for clinical meet- ings where images ar e shown on a television. Recent develop- ments have blurred the boundaries of there two types. Origi- nally intended for use as a presentation tool, lower-priced ana- logue cameras combine all the visual information into a single video signal (composite video). Compared to a digital camera the quality is inferior; however, coupled with a video-capture unit (‘imagegrabber’), it is a low-cost alternative. Video cap- ture units accept analogue signals and ‘capture’ images at reso- lutions of up to 1500 x 1125 pixels through a technique based on oversampling the analogue signal. The resolutions claimed are spurious and those above the base camera r esolution result in the magnification of electronic artefacts. There are several cameras that are able to provide both func- tions. A high-end example is the Kontron/ Jenoptik ProgRes 3012 that is a dedicated microscopy camera capable of captur- ing still images up to 20 megapixels in size (5000 x 4000 pixels) and transmitting them via a digital connection to a computer. However, it also is capable of composite analogue video out- put and can be used for live image display. Similarly, the Nikon Coolpix cameras (950/990/995/5000/5700) are general con- sumer, still image cameras that store images up to 5 megapixels in size onto removable compact flash media. They also have composite video output that provides a live image view that is updated at 15 frames per second. Dual still-image with live video output is slowly becoming the norm because even when camera and eyepieces are parfocal, it is still preferable for fine focusing to be performed automati- cally by the camera rather than through microscope eyepieces. The only way to provide this function is through some form of ‘live’ or rapidly updated camera view as found in the Nikon DXM1200, Kontron/ Jenoptik Prog Res 3012 or C14 that pro- vide this within its image-grabbing software. Table 1: Some typical digital microscopy cameras High-end (>US$5000) Mid-range (US$1000-5000) Cameras which may optionally be microscope-mounted (US$500-5000) Nikon DXM1200 ProgRes C14 Nikon DN100 Leica DC150, DC300 Olympus Camedia E-10, C -4040 Leica DC500 Zeiss Axiocam HR Olympus DP12, DP50 Zeiss Axiocam MR Nikon Coolpix 950, 990, 4500, 5000Nikon D100, D1, D1H, D1XOther digital SLRs Note: This list is by no means comprehensive and represents the more readily available models through microscope vendors Leong et al: Digital photography in anatomical pathology 64 J Postgrad Med March 2004 Vol 50 Issue 1� � 64 CMYK ‘Live video’ digital cameras remain in demand for applications where real-time image processing is required, such as in ro- botic telepathology where the live camera view is used for con- sultation or as the means by which the softwar e-based autofocus is calculated. This area continues to develop with manufacturers using multiple image sensors to provide better image quality (so called ‘triple-chip CCDs’) and/or avoiding analogue artefacts through use of the relatively new IEEE1394 (‘Firewire’) digital interface. The Sony DFW series is an ex- ample of a range of microscope cameras that provide a pure digital live image of up to 1280 x 1024 pixels at 15 frames/ second through an IEEE1394 interface. Macroscopic Photography In the past, the reliance has been on chemical film photogra- phy to produce black and white or more expensive colour prints to supplement the descriptive prose of the pathology report. The photocopier has also been employed for some specimens, particularly those with a flat cut surface, to produce instant, inexpensive prints that could be used to mark sites of block sampling and as a replacement for line drawings. Photomicro- graphs did not accompany such reports and wer e reserved for clinico-pathological conferences and published text. Digital photography has changed the face of the pathology report significantly, allowing the incorporation of coloured prints of the gross specimen as well of relevant microscopic features; greatly enhancing the accuracy and reproducibility of the findings. As there ar e essentially no running costs in- volved, there is no limit to the number of images used so that it is possible to replace or supplement traditional, often lengthy, descriptive prose with high quality images that greatly improve the quality of the report. Coupled with synoptic texts, r eports are more accurate and succinct3 and lengthy block keys and hand-drawn diagrams can be replaced with digital photographs. Several reports describe home-assembled digital photographic set-ups for routine use in surgical and autopsy pathology. 4-7 For an optimal set-up of gross photography the camera must be controlled remotely to avoid contamination, and downloading of images and specimen/accession number rec- ognition should be automatic. While it is possible to obtain good digital images with many of the lower end cameras that produce 3-5 megapixel images, automatic downloading and remote control are much more difficult to achieve. One group has successfully developed software for these two functions employing the Coolpix 995 camera, spending about US$26500 on the system and achieving an annual savings of US$7500 in labour and film costs.8 We have pr eviously described a prototype of an automated system,9 which is now replaced by the more sophisticated MacroPath system (Milestone, Bergamo, Italy, retails at about US$18000). This system allows remote control of a video cam- era (with imagegrabber) through a foot pedal for zoom, cap- ture and save featur es (Figure 1), auto focus, automatic downloading and the use and automatic recognition of spe- cific specimen accession numbers. In addition, the requisition form may be photographed and stored, and the images can be stored to the laboratory information system, shared on the Intranet among pathologists or via floppy disk or CD-ROM. A useful feature of the system is the provision for measurements to be done on formalin-resistant screen. Prior calibration al- lows measurements to be made by drawing a line between points of interest on the image with a stylus or even finger. In a similar manner sites of sampling can be indicated with auto- matic labelling so that the use of lengthy conventional block keys and important aspect of report documentation can be dispensed with (Figure 2). Selected images can be incorpo- Leong et al: Digital photography in anatomical pathology Figure 1: The Milestone MacroPath system comprising a camera (black arrow), foot pedal control with zoom in, zoom out and freeze pedals (white arrow), formalin-resistant touch screen monitor with attached splash-proof computer (yellow ar row), infra-red bar code scanner (red arrow) and keyboard (broad arrow) Figure 2: Digital image of uterus, bilateral tubes and ovaries with leiomyomas including a large subserosal leiomyoma. Measurements of the specimen made on the image and sites of sampling (block key) are shown. When supplemented by images of various cut surfaces, descriptive prose and block keys can be dispensed with. 65J Postgrad Med March 2004 Vol 50 Issue 1 � � C MYK 65 rated into the report or exported through the Internet or ap- pended as colour prints. Microscopic Photography (Photomicrography) A common mistake when making the transition to digital imaging is to be fixated upon the camera. Although it is the camera that converts the analogue image into a digital form, it is the interaction of all the components that determines the final outcome. Therefore, any weak component in the imaging chain can be detrimental to the final product. The minimum hardware components requir ed for digital photomicrography are a digital camera, camera-to-microscope adaptor, micro- scope, and a computer with an adequate graphics card and monitor. Factors such as system Central Processing Unit speed, hard drive speed and capacity, and the nature of the remov- able storage medium are secondary concerns. Dedicated digital microscopy cameras generally do not have a lens and attach to the microscope via a common connection, the ‘C-mount’, a 1-inch diameter x 32-threads/inch screw origi- nally designed for 16-mm motion video cameras but now also adopted for lens- less video cameras. When a digital camera has a fixed lens, a feature of general consumer cameras such as the Nikon Coolpix series and Olympus D-3030 series, a ‘C- mount adaptor’ provides another layer of optics to compen- sate for the camera lens.8 The importance of good optics cannot be overemphasized, ir- respective of the image sensor used. Noticeable improvement in digital image quality is achieved by upgrading objective lenses from fluorochromatic (‘plan fluor’) to apochromatic (‘plan apo’). It is beyond the scope of this paper to discuss the relationship between pixel size and optical resolution. Suffice to say, each component in the imaging pathway should be cali- brated to its optimum state. This includes Köhler illumina- tion for the microscope, white balancing for the camera, and monitor calibration (contrast, brightness and gamma settings). Digitisation of Previous Collections of Projection Slides (K odachromes) The conversion to digital images makes previous collections of macroscopic and microscopic images on film obsolete. These can be scanned with a film scanner (if the chemical images are still in strips they can be scanned more expediently) and stored as digital images, but a more convenient and expedient method is to photograph the projection slide over an X-ray viewing box, a procedure that takes no more than a few seconds compar ed to the more laborious and lengthy scanning process. Some editing of the acquired image may be necessary. Similarly, the difficulties of acquiring low-power scanning images of an en- tire tissue section can be circumvented by photographing the tissue section over an X-ray viewing box. Practical Aspects Instead of mounting the camera via the C-mount it may be placed against one eyepiece of the microscope and images can be captured through this lens. This method is cumbersome and requir es a steady hand. As previously discussed, it is pref- erable to allow the camera to perform automatic fine focussing rather than to focus through the microscope eyepiece even if the system is parfocal. Although one axiom with digital cam- eras is “ what you see is the image you capture”, uneven field illumination or vignetting may commonly be present in low- power images. While the view through the eyepieces may not reveal unevenness in the peripheral field, as the human eye may not detect it, the problem is clearly displayed in the dig- ital image. Vignetting is the r esult of the coning of the light path between the microscope objective and the camera or may be due to an uneven light source. The pre-digital approach was simply to crop the image, decreasing the effective field. With digital imaging, retrospective correction of uneven illu- mination is possible through mathematical modelling of the illumination pattern.10 An even simpler method is to tak e a blank reference image and subtract this from subsequent im- ages. This reference frame corrects any uneven illumination present.11 Vignetting can also be avoided to a large extent by using higher magnification objectives. Most cameras provide both optical and digital zoom facilities. While optical zoom is useful, digital zoom should be avoided as the image tends to blur and the magnification is not true. Simple digital magnification or increasing the number of pix els in an image will not provide increased resolvable detail be- yond the limits of the optics. Three main variables influence the quality of a digital microscopy image. A combination of the objective lens, image sensor size, and the light path between the objective and the sensor determines the physical area imaged in a single field. As with the optical microscope, higher numerical apertures are associated with higher-grade objective lenses and there- fore better quality digital images.1 The numerical aperture, quality, and type of the microscope objective lens determines the accuracy in focusing light on the imaging sensor and there- fore image quality and resolution. The image sensor in the digital camera determines how this light is converted into a digital signal and the number of pixels created. Digital microscopy images are acquired for various purposes including teaching and clinical meetings, conference presen- tations, diagnosis, image analysis and archiving, each with slightly different requirements. Images for a clinical meeting are best projected live, in real-time, with no delays, and the image size requir ements would be modest as the commonest mode of display is an analogue television set that has, at best, 550-600 horizontal lines. Thus, a 640 x 480 pixel image would suffice. On the other hand, an image for publication would require sufficient pixels to satisfy the publisher. A 4 x 3 inch image on a 300 dpi home printer would ideally require at least a 1200 x 900 pixel image. In PowerPoint presentations only 72 pixels per inch are required. For telepathology purposes it seems that the price and avail- ability of digital imaging equipment rather than scientific con- siderations of adequate sampling have guided pathologists. Leong et al: Digital photography in anatomical pathology 66 J Postgrad Med March 2004 Vol 50 Issue 1� � 66 CMYK Many studies have employed digital cameras with analogue video output. Such cameras typically have an image dimen- sion in the vicinity of 768 x 576 pixels. Thus, some patholo- gists have based their clinical practice not upon a scientific measure but on an arbitrary figure set by camera manufactur- ers. Telepathology systems mostly employ images that are ei- ther in the region of 352 x 255 (video-conferencing size) or 768 x 576 pixels (typical framegrabber image size). Image Storage There are two aspects for consideration, the physical storage and the method of storage. Physical storage for daily work is commonly on fixed hard drives, which currently vary between 20-100 gigabytes in size. Portable hard drives offering similar performance through a Firewire (IEEE1394) or USB2 connec- tion are increasingly popular; however, avoid the comparatively slower USB1 connections. Removable media such as Zip disks and optical disks are relatively expensive and USB media allow a cheap and convenient method of transferring data between computers. CD-R (650-700M) is the most economical form of near-permanent storage although DVD-R (4.6G) should become cost-effective in the near future. Large image datasets are the norm when dealing with digi- tised slides and some form of data compression is often neces- sary. It is also necessary to compress images when the trans- mission bandwidth is narrow. The compromise for a smaller file is a decrease in quality. To the human eye, this may not be noticeable but poor quality digital compression such as JPEG or JPEG2000 inevitably introduces digital artefacts and alters the numerical values of the image. This may be significant if the image is to be used for image analysis. For printing, view- ing and even interpretation, poor quality compressed images are acceptable provided the compression ratio is controlled. Image Printing Sales of general consumer digital cameras now exceed that of film cameras so that many film processing laboratories will now print digital images on glossy photo paper that are indis- tinguishable from those derived from 35-mm film. Submis- sion of digital images for printing via the Internet is an accept- able and convenient route provided there is access to a good Internet connection and often, because of the size of the data, Broadband connection may be required. Inkjet printing, when coupled with the appropriate paper can produce publication-quality results and it is not uncommon for professional photographers to exhibit their work in this manner. Colour laser printers ar e better suited for high vol- ume output but are generally not appropriate for photographic- quality prints. The relationship between ink, paper and envi- ronment determines the quality and longevity of the print. Inkjet prints fade or show colour alteration, a phenomenon that usually takes weeks to months and is less lik ely to occur if the print is kept away from direct sunlight and behind glass, protected from environmental elements. ‘Curing’ the print immediately after production by placing it away from air and light, e.g. in an envelope, for a period of time, has been a sug- gested solution but there is no evidence to show that this im- proves subsequent colour stability. Matt photographic rather than glossy paper improves longevity but the more-familiar glossy finish may be preferred. Production of monochromatic images is often necessary as colour plate publication is expensive. Until r ecently, many photo printers were unable to print black-and-white images without introducing some unsightly colour tint. A simple so- lution was to switch off the colour cartridge and use only black ink; however, the results were often slightly coarse particularly across smooth monochromatic tonal gradients. Recent gen- eration photo printers such as the Epson 1270 are able to pro- duce a neutral grey using a balanced mixture of black and col- our inks. Even so, many prefer the greater control and tonal range offered by dedicated monochrome ink cartridges. If the publisher accepts digital files, sending an uncompressed file (e.g., TIFF format, sRGB colorspace) on CD-ROM is al- ways preferable to their attempt at scanning the printed im- age. This circumvents the need to print. (Although printed illustrations are still required in the initial review process they need not be of the best quality.) Total Slide Digitisation and Archiving The traditional glass slide collection of interesting cases re- mains an excellent teaching resource provided the collection is maintained. Unfortunately, such collections are fragile, will fade, cannot be replaced if damaged, are difficult to transport, and cannot be readily shared among many and never simulta- neously. Additionally, they are not in a form for immediate presentation or publication. Digital image archiving provides the solution to many of these drawbacks. At the very least, multiple representative images may be stored and, provided they are properly catalogued, can be retrieved and dissemi- nated far more easily and widely, and without fear of damage. Sampling error and inadequacy of representative images are overcome by the development of total slide digitisation.12 Sys- tems from companies like Zem/Nikon, 13 BacusLabs,14 Aperio, 15 and Interscope16 can scan the entire tissue area on a slide at high-power (20x or 40x objective), and present the data in a manner that replicates the actual slide without loss of resolu- tion. Such systems may form the basis of anatomical pathol- ogy laboratories of the future. When coupled with rapid tis- sue processing systems such as the microwave-stimulated sys- tem from Milestone,17 which has the capability of ultra-rapid tissue processing from fresh tissue to paraffin-embedded block within 30-90 minutes, it will allow large processing “factories” to service pathologists located at distant sites. Specimens can be transported to the centralised “factory” where examination, photography of macroscopic specimens, ultra-rapid tissue processing and whole slide digitisation takes place. The im- ages of gross specimens and total digitised scans of microscopic sections can be transmitted for reporting by pathologists lo- cated at distant laboratories, eliminating the need to incur the expense of maintaining such laboratories. 18 Leong et al: Digital photography in anatomical pathology 67J Postgrad Med March 2004 Vol 50 Issue 1 � � C MYK 67 The American Board of Pathology conducts its exam with vir- tual microscopic slides acquired through whole slide digitisation employing the BacusLabs system.14 Telepathology Telepathology is practised to varying degr ees worldwide and is primarily used for consultation.19 This may be for primary opin- ion, frozen section consultation being an important applica- tion in Europe 20-22 or for second opinion.23,24 There is some confusion in the terminology of the current methodologies available. “Simple store-and-forward” telepathology is the act of capturing several representative images beforehand and for- war ding them to a remote site for review. 19,25 “Dynamic” telepathology has been used in the United States to describe real-time consultation where the pathologist views the live microscope image updated at least 15 frames per second.26 This is commonly done through a video-conferencing codec.27 How- ever, elsewhere, users have adopted the term “dynamic” to describe any consultation occurring in real-time irrespective of the mode of image transfer. 2 8 In between these two modalities the terminology overlaps and other terms such as ‘active’ and ‘passive’21,29 have been introduced in an attempt to provide more clarity. The introduction of digital slide scanning systems has further complicated terminology.29,30 They can be regarded as extremely sophisticated store-and-forward systems but the difference between whole slide digitisation and a dozen representative pictures transmitted by email is considerable. We have sug- gested a classification for such telepathology systems.2 In recent times, reference immunohistology laboratories in the United States31 and Australia32 have employed the Internet to post images of stains performed on referred cases in order the circumvent the delays in courier/postal deliveries of the stained glass slides. While these can currently only be of selected rep- resentative fields, the advent of total slide digitisation will open new avenues of rapid slide transfer and examination. Quality Assurance and Educational Programmes The use of representative digital images or entire digitised slides will eliminate the need to section and disseminate multiple glass slides, besides ensuring that all recipients see exactly the same section.33,34 This circumvents the time and costs involved for slide distribution and is particularly important in cytology quality assurance programmes where only limited smears are available.25 Digital images will also eliminate the need for cum- bersome storage of educational and quality assurance glass slides that fade with time and take up space. Furthermore, digital images can be edited and improved in quality and can be made available on websites for educational purposes after completion of the quality assurance exercise.25,34 The same advantages apply to both— digital microscopic images and images of gross specimens from autopsy or surgical biopsy ex- amination.9 Digital images circulated on CD-ROM will also gradually re- place the traditional glass slides that are used for the teaching slide seminar, and colour images can also be made available as a component of the handout for such seminars and lectures. Digital Pathology Images in Education The idea of using image databases as educational resources is not new. There ar e several websites that provide pathology- related images, the most well-known being Webpath.35 How- ever, the quality of the images is often poor, both in diagnostic sophistication and quality, and in the number of macroscopic and histological images. A major problem has been the lack of standard in terms of image size. Many older databases began collating images before high-quality digital cameras were read- ily available. Total slide digitisation now replaces the need to store representative images of each teaching case as they vir- tually contain all the data in the tissue section. At a macroscopic level, three-dimensional digital simulations of organs provide an alternative to plastination or formalin- filled pathology pots. Digital representations are not subject to the legalities of organ and tissue retention; neither do they require maintenance or physical space for storage. There ar e several methods of creating a three-dimensional specimen. Contemporary rendering software has capabilities of recon- structing faces using only two different views. Rendering dis- eased human organs is a more difficult task due to the more complex shapes and textures. At Oxford University, we used a lower cost but more labour-intensive method of photograph- ing the organ from at least 12 angles and reconstructing it into an interactive rotatable object, deliverable through a web in- terface as a Quicktime VR object.36 Importantly, both gross and microscopic digital images allow a greater scope for inter- active teaching programmes that can be posted on websites or distributed on CD-ROM. At a microscopic level, visualisation of histological image data in three dimensions does not significantly improve diagnostic accuracy but can be a useful exercise in understanding the micro-anatomical growth and arrangement of structures. For example, the impact and influence a tumour has on its sur- rounding microvasculature may provide clues to the diagnosis and understanding of its growth pattern. Image data derived from confocal microscopy can be rendered into a three-dimen- sional image by computer software. It is also possible to do this at a lower cost using serial sections and normal microscopy images, as we have done with tubular carcinoma of the breast.37 Medical Vision Systems Automated cytology screening machines have made major in- roads into pathology practice where they address the shortage of cytology screeners and reduce the false-negative detection rate of the human screener. 38,39 It is unlikely that computer di- agnosis will completely replace human interpretation. How- ever, computer-assisted systems can enhance human perform- ance by identifying highly suspicious areas or cells in cytologi- cal preparations, thus acting as a safety net or quality assur- ance for the pathologist. We are engaged in the preliminary Leong et al: Digital photography in anatomical pathology 68 J Postgrad Med March 2004 Vol 50 Issue 1� � 68 CMYK development of such systems.37 The day may come when it will be regarded as negligent in a court of law if the reporting pathologist does not employ such a system for back-up or qual- ity assurance. Fortunately, as yet, there are no such systems available. The attempt to teach machines to recognise histological con- ditions is a worthy pursuit with the intermediate benefit of the objectification of histological interpretation. Given an ap- propriate training set, it is possible to derive objective criteria to describe a histological condition and it is also possible to statistically rank the features in terms of diagnostic strength.37 Work in several areas including colonic dysplasia,40 grading of cervical intraepithelial neoplasia,41 endometrial hyperplasia,4 2 and prostatic biopsy diagnosis43,44 has made this a reality. Other Applications of Digital Imaging in Pathology Quantitative pathology may be viewed as an instrument of research. Its penetration into daily clinical practice is low, de- spite the existence of many aspects of histopathology, which would benefit from some form of quantification— mitotic counting, assessment of tumour volume and extent, morphom- etry and immunohistological densitometric quantitation be- ing some examples. Quantitation of digital images will be much more reliable than assessment by eyeballing, using poorly de- fined criteria that cannot be reliably reproduced. Image analy- sis may render tumour grading to be more accurate and r epro- ducible, and automation will make quantitation more accept- able in routine practice. Some critics of digital imaging cite the technique as an invita- tion to scientific fraud.45 In fact, the opportunity to be fraudu- lent existed long before the advent of digital imaging and there are certainly easier ways of falsifying scientific data. It is im- possible to have an image that is completely free of some form of processing. For single-chip CCD cameras, the presence of a colour filter array means that there is comple x data interpola- tion being employed to create the final image. Additionally, there is further processing of image data to improve the signal to noise ratio. Many techniques such as spectral imaging and confocal microscopy are dependent upon processing of the image data to produce a viewable image. Photography of immunofluoresence is not optimal without some degree of contrast and brightness adjustment afterwards in an attempt to have the digital image match what was viewed through the microscope eyepieces. The closest equivalent to a raw nega- tive are the unprocessed data files that some cameras such as the Nikon D1 series and Coolpix 5700 can produce. This is image data in a proprietary format that requires software to assemble it into a digital image. In some situations, image processing can assist interpretation. We all are aware of inter- and intra-laboratory variations in histochemical and immunohistological staining. The human eye adapts and compensates for these variations. However, when sections fade with time or sun-exposure, r e-staining is often necessary. With digital image processing it is possible to rescale the distribution of values in a faded image to provide clarity equivalent to re-staining the section. Conclusion While digital imaging is relatively new in pathology, it is rap- idly being adopted in many areas because of cost savings, ra- pidity and convenience of usage, and the many new applica- tions it offers. The use of digital images enhances the quality and accuracy of pathology reports and whole slide digitisation has the potential to completely change the way pathologists examine and store tissue sections. In addition, the applica- tions in education are numerous and include conference pres- entations, publication, undergraduate and postgraduate teach- ing as well as remote consultation through telepathology, and quality assurance programmes. Digital images also have many other applications that are yet to be exploited including morphometric and densitometric quantitation and in medi- cal vision systems with the ultimate aim of computer-gener- ated histological diagnosis, either as a primary diagnostic tool, as a back-up system to aid the pathologist in diagnosis, or in quality control. Rapidly falling costs and extensive applications makes the adoption of digital imaging in anatomical pathol- ogy laboratories an essential consideration. References 1. Leong AS, James C, Thomas A. Handbook of surgical pathology. London: Church- ill Livingstone; 1996. p. 43-100, 177-92. 2. Leong FJ, Leong AS. Digital imaging applications in anatomical pathology. Adv Anat Path 2003;10:88-95. 3. Leong AS. Synoptic/checklist reporting of breast biopsies: Has the time come? Breast J 2001;7:271-4. 4. Schubert E, Gross W, Sideritis RH, Deckenbaugh L , He F, Becich MJ. A patholo- gist-designed imaging system for anatomic pathology signout, teaching and re- search. Sem Diagn Pathol 1994;11:263-73. 5. Cruz D, Seixas MA. A surgical pathology system for gross specimen examination. Proc AMIA Symp 1999;236-40. 6. Belanger AJ, L opes AE, Sinard JH. Implementation of a practical digital imaging system for routine gross photography in an autopsy environment. Arch Pathol Lab Med 2000;124:160-5. 7. Marchevsky AM, Dulbandzhyan R, Seely K, Carey S, Duncan RG. Storage and dis- tribution of pathology digital images using integrated Web-based viewing systems. Arch Pathol Lab Med 2002;126:533-9. 8. Park R, Eom J, Park P, Lee K, Joo H. Automation of gross photography using a remote-controlled digital camera system. Arch Pathol Lab Med 2003;127:726-31. 9. Leong AS, Visinoni F, Visinoni C, Milios J. An advanced digital image-capture com- puter system for gross specimens: A substitute for gross description. Pathology 2000;32:131-5. 10. Likar B, Maintz JBA, Viergever MA, Pernus F. R etrospective shading correction based on entropy minimization. J Microsc 2000;197:285-95. 11. Leong FJ, Brady M, McGee JO. Correction of uneven illumination (vignetting) in digital microscopy images. J Clin Pathol 2003;56:619-21. 12. Leong FJ, McGee JO. Automated complete slide digitization: a medium for simul- taneous viewing by multiple pathologists. J Pathol 2001;195:508-14. 13. N i kon UK. http://www. n i k on.co.uk and http://www. n i kon.co.uk/inst/ flashframeset.htm. Accessed September 2003. 14. BacusLabs. http://www.bacuslabs.com/. Accessed September 2003. 15. Aperio Technologies. http://www.aperio.com/home.asp. Accessed September 2003. 16. Interscope. http://www.interscopetech.com/index-products.htm. Accessed Sep- tember 2003. 17. Visinoni F, Milios J, LEONG AS -Y, Boon ME, Kok LP. Ultra rapid microwave/variable pressure-induced tissue processing-description of a new tissue processor. J Histotechol 1988;21:219-24. 18. Leong AS -Y, Leong FJW-M. Telepathology-the system for Australia. National Telepathology Conference, Sydney, October 5-6, 2001. 19. Leong FJW-M, Graham AK, Gahm T, McGee JOD. Telepathology: Clinical utility and methodology. In : Underwood J, ed. Recent Advances in Histopathology. Lon- don: Churchill Livingstone;1999. p. 217-40. 20. Eide TJ, Nordrum I. Frozen section service via the telenetwork in Northern Norway. Zentralbl Pathol 1992;138:409-12. 21. Kayser K. Telepathology in Europe. Its practical use. Arch Anat Cytol Pathol 1995;43:196-9. 22. Wellnitz U, Fritz P, Voudouri V, Linder A, Toomes H, Schmid J et al. The validity of telepathological frozen section diagnosis with ISDN- mediated remote microscopy. Leong et al: Digital photography in anatomical pathology 69J Postgrad Med March 2004 Vol 50 Issue 1 � � C MYK 69 Virchows Arch 2000;437:52-7. 23. Milosavljevic II, Spasic P, Mihailovic D, Kostov M, Mijuskovic S, Mijatovic D, et al. Telepathology-second opinion network in Yugoslavia. Adv Clin Path 1998;2:156. 24. Della Mea V, Beltrami CA. Telepathology applications of the Internet multimedia electronic mail. Med Inform (Lond) 1998;23:237-44. 25. Lee ES, Kim IS, Choi JS , Yeom B, Kim HK, Ahn GH, et al. Practical telepathology using a digital camera and the Internet. Telemed J E Health 2002;8:159-65. 26. Dunn BE, Choi H , Almagro UA, Recla DL. Combined robotic and nonrobotic telepathology as an integral service component of a geographically dispersed labo- ratory network. Hum Pathol 2001;32:1300-3. 27. Vazir MH, Loane MA, W ootton R. A pilot study of low-cost dynamic telepathology using the public telephone network. J Telemed Telecare 1998;4:168-71. 28. Afework A, Beynon MD, Bustamante F, Cho S, Demarzo A, Ferreira R, et al. Digital dynamic telepathology—the Virtual Microscope. Proc AMIA Symp 1998:912-6. 29. Kayser K, Syzmas J, Weinstein R. Telepathology-Telecommunication, electronic education and publication in pathology. Berlin: Springer-Verlag; 1999. 30. Demichelis F, Barbareschi M, Dalla Palma P, Forti S. The virtual case: a new method to completely digitize cytological and histological slides. Virchows Arch 2002;441:159-64. 31. PhenoPath Laboratories. http://www.phenopath.com/. Accessed September 2003. 32. AIP Laboratory Zoom-in system. http://www.aiplab.com/online/login.mv [Accessed February 2003]. 33. Leong FJ, Graham AK , Schwarzmann P, McGee JO. Clinical trial of telepathology as an alternative modality in breast histopathology quality assurance. Telemed J E Health 2000;6:373-7. 34. Leong FJ, McGee JO. Robotic interactive telepathology in proficiency testing/quality assurance schemes. Electr J Pathol Histol 2001;012-01. 35. Webpath. http://www -medlib.med.utah.edu/WebPath/webpath.html. Accessed October 2003 36. Oxford University Laboratory Medicine Course. http://www.ndcls.ox.ac.uk/LMC. Accessed September 2003. 37. Leong FJ . An automated diagnostic system for tubular carcinoma of the breast- a model for computer-based cancer diagnosis. DPhil Thesis: Oxford Univer- sity;2002. 38. Cenci M, Nagar C, Giovagnoli MR, Vecchione A. The PAPNET system for quality control of cervical smears: validation and limits. Anticancer Res 1997;17:4731-4. 39. Patten SF Jr, Lee JS, Wilbur DC, Bonfiglio TA, Colgan TJ, Richart RM, et al. The AutoPap 300 QC System multi-center clinical trials for use in quality control re- screening of cervical smears: II. A prospective intended use study. Cancer 1997;81:337-42. 40. Hamilton PW, Bartels PH, Thompson D, Anderson NH, Montironi R, Sloan JM. Au- tomated location of dysplastic fields in colorectal histology using image texture analysis. J Pathol 1997;182:68-75. 41. Keenan SJ, Diamond J, Glenn McCluggage W, Bharucha H, Thompson D, Bartels PH et al. An automated machine vision system for the histological grading of cer- vical intraepithelial neoplasia (CIN). J Pathol 2000;192:351-62. 42. Morrison ML , McCluggage WG, Price GJ, Diamond J, Sheeran MR, Mulholland KM, et al. Expert system support using a Bayesian belief network for the classifi- cation of endometrial hyperplasia. J Pathol 2002;197:403-14. 43. Bartels PH, Thompson D, Bartels HG, Montironi R, Scarpelli M, Hamilton PW. Ma- chine vision-based histometry of premalignant and malignant prostatic lesions. Pathol Res Pract 1995;191:935-44. 44. Montironi R, Bartels PH, Hamilton PW, Thompson D. Atypical adenomatous hy- perplasia (adenosis) of the prostate:development of a Bayesian belief network for its distinction from well-differentiated adenocarcinoma. Hum Pathol 1996;27: 396-407. 45. Suvarna SK, Ansary MA. Histopathology and the ‘third great lie’. When is an image not a scientifically authentic image? Histopathology 2001;39:441-6. Leong et al: Digital photography in anatomical pathology