Data diversity and virtual imaging in AI-based diagnosis: A case study based on COVID-19

• DOI: 10.48550/arxiv.2308.09730
• Publication Date: 2023-01-01
• Authors: Fakrul Islam Tushar; Lavsen Dahal; Saman Sotoudeh‐Paima; Ehsan Abadi; W. Paul Segars; Ehsan Samei; Joseph Y. Lo
• Keywords: FOS: Computer and information sciences; Computer Science - Machine Learning; Image and Video Processing (eess.IV); FOS: Electrical engineering, electronic engineering, information engineering; Electrical Engineering and Systems Science - Image and Video Processing; 3. Good health; Machine Learning (cs.LG)

The credibility of AI models in medical imaging is often challenged by reproducibility issues and obscured clinical insights, a reality highlighted during the COVID-19 pandemic many reports near-perfect artificial intelligence (AI) that all failed to generalize. To address these concerns, we propose virtual trial framework, employing diverse collection images are both simulated. In this study, serves as case example unveil intrinsic extrinsic factors influencing performance. Our findings underscore significant impact dataset characteristics on efficacy. Even when trained large, datasets with thousands patients, performance plummeted up 20% generalization. However, trials offer robust platform for objective assessment, unveiling nuanced insights into relationships between patient- physics-based For instance, disease extent markedly influenced efficacy, computed tomography (CT) out-performed chest radiography (CXR), while dose exhibited minimal impact. Using study verified radiology suffer from crisis. Virtual not only offered solution assessment but also extracted several insights. This illuminates path leveraging augment reliability, transparency, relevance imaging.