key: cord-0825111-pb4921w8 authors: Pirgulov, S. title: 103P Effectiveness of artificial intelligence in retrospective COVID-19 lung CT analysis for lung cancer detection date: 2022-04-30 journal: Annals of Oncology DOI: 10.1016/j.annonc.2022.02.129 sha: 92df045324864999031b6872aaa46d279402acbf doc_id: 825111 cord_uid: pb4921w8 nan Background: With the increasing use of the computed tomography (CT) scans, the clinicians more commonly encounter incidentally detected pulmonary nodules, parts of which are finally diagnosed as primary lung cancer during follow-up. The clinical characteristics of incidentally detected lung cancers (IDLCs) have not been well known in South Korea. In this study, we compared the clinical characteristics and prognosis of IDLC and screening detected lung cancer (SDLC). Methods: This retrospective study included the subjects with pulmonary nodules (<3cm) at the baseline CT scans, which were pathologically confirmed as primary lung cancer in year 2015. The study population was classified as IDLC and SDLC according to the setting of the first pulmonary nodule detection. The symptomatic subjects at the time of the pulmonary nodule detection were excluded. Clinicoradiologic characteristics and overall survival (OS) rates were compared between the IDLC and SDLC groups. Conclusions: Among non-symptomatic lung cancer patients, the proportion of IDLC was about 38% in our cohort. The IDLC group was associated with old age, smokers, history of other malignancy, and non-adenocarcinoma histology. However, the prognosis of the IDLC group was not inferior to the SDLC group. Legal entity responsible for the study: The authors. Funding: Has not received any funding. Background: In the context of the Covid-19 pandemic, due to the dramatically increased volume of CT exams and COVID-19 pneumonia signs as a background there is a the risk of missing nodules. The goal of our project was to evaluate the effectiveness of using automated image processing with AI in chest COVID CT scans analysis for pulmonary nodes detection. Methods: 9035 scans were selected for the study. The inclusion criteria were age over 45 years and the volume of lung affected by COVID less than 50%. The depersonalized images retrospectively were processed on the Botkin.AI platform. The identified nodules were automatically marked, countered and classified according to Lung-RADS criteria. After the automatic AI analysis, the obtained results were assess by experts to check the correctness of the data. Results: As a result of AI analysis absence of nodules was confirmed in 8123 (89.9%) of cases. Pathological nodules were detected in 912 (10.1%) of cases: Lung-RADS 2, 3 -662 pts (68% of them and 7% of the total number of cases); Lung-RADS 4a, 4b -290 pts (32% of cases with nodules and 3% of the total number of processed studies). The results of experts reassessment were: -nodules were confirmed in 139 cases (48% of the number of nodal changes of the Lung-RADS 4a, 4b category detected by AI);nodal formations were not confirmed in 132 cases (45% of the number of nodal changes of the Lung-RADS 4a, 4b category detected by AI); -19 cases were classified as doubtful (7% of the number of nodal changes identified by AI in the Lung-RADS 4a, 4b category). 27 cases, with the nodules detected by AI were missed by radiologists. Conclusions: Artificial intelligence can help to decrease the healthcare system overload and optimize the diagnostics of lung nodules. Thanks to artificial intelligence, the volume of research for the retrospective experts assessment of significant changes in the lungs belonging to the Lung-RADS 4a, 4c category is 3.2% of the total research volume. Automated image processing algorithms (Botkin.AI platform) provides an opportunity to reliably identify or exclude lung nodules against the background of inflammatory changes caused by the "new coronavirus infection". Legal entity responsible for the study: The author. Funding: Has not received any funding. Disclosure: The author has declared no conflicts of interest. https://doi.org/10.1016/j.annonc.2022.02.129 The potential for lung cancer detection in COVID-19 CT scans with AI technologies usage Oncologi, Nizhny Novgorod Regional Clinical Oncology Center, Nizhny Novgorod, Russian Federation Background: The COVID-19 pandemic has made the patient journey very difficult especially for diagnostics of new lung cancer (LC) cases because of lockdown, social distancing, similarity of symptoms and limitations with healthcare access. At the same time thousands of patients underwent CT for COVID detection. The aim of our study was to assess AI technology for LC detection in a COVID CT scan database. Methods: Chest CT scans (without age, sex, smoking history, COVID-19 severity grade and other limitations) were retrospectively anonymized and analyzed by AI platform (BotkinAI). All findings were classified according to Lung-RADS criteria, reassessed by experts, patients were checked with regional onco registry and, if necessary, followed up for LC confirmation. The COVID-19 pandemic gave a unique chance for incidental pulmonary nodes detection and artificial intelligent technology can provide support to physicians, save time and increase effectiveness of CT scans analysis. Legal entity responsible for the study: The author. Funding: AstraZeneca. Oncologic Radiology Unit Thoracic Surgery Unit, Department of Cardiac, Thoracic and Vascular Sciences Background: Radiomics is a quantitative approach to medical imaging consisting in the extraction of a large number of features (fts) from diagnostic images that can be associated with tumour pathophysiology and converted into mineable high-dimensional data. We define a model of workflow for extraction and selection of radiomics fts from the baseline computed tomography scan images of stage III NSCLC pts, thus creating a radiomics profile to guide the clinical decision-making process. We retrospectively collected data of stage III NSCLC pts referred to Veneto Institute of Oncology from 2012 to 2021. The radiomics pipeline includes (1) the definition of inclusion/exclusion criteria based on image quality and clinico-pathological data, (2) data selection for training and validation cohorts, (3) image Volume 33 -Issue S2 -2022 S79