Reality Check, In Silico COVID-19 Imaging Case-study (ReviCOVID)
Virtual Imaging Trials Improved the Transparency and Reliability of AI Systems
Project Page: https://fitushar.github.io/ReviCOVID.github.io/
This project uses COVID-19 as a case study to explore the impact of data diversity and virtual imaging on AI diagnostic accuracy. By analyzing clinical and virtual CT and chest X-ray (CXR) images, we investigate how variations in dataset characteristics—such as imaging type and disease severity—affect AI performance. Virtual imaging is employed as a tool to evaluate model reliability under diverse conditions. The findings will provide insights into designing more robust AI models, with applications extending to other diseases beyond COVID-19, enhancing diagnostic consistency across varied patient data.
This project uses COVID-19 as a case study to explore the impact of data diversity and virtual imaging on AI diagnostic accuracy. By analyzing clinical and virtual CT and chest X-ray (CXR) images, we investigate how variations in dataset characteristics—such as imaging type and disease severity—affect AI performance. Virtual imaging is employed as a tool to evaluate model reliability under diverse conditions. The findings will provide insights into designing more robust AI models, with applications extending to other diseases beyond COVID-19, enhancing diagnostic consistency across varied patient data.