Lunit, a leading provider of AI-powered cancer diagnostics solutions, has recently published a study in the renowned medical imaging journal ‘Radiology.’ The study, conducted by Seoul National University Hospital, examined the impact of medical AI solutions’ accuracy on radiologists’ diagnostic determination. The findings reveal that AI solutions with high diagnostic accuracy significantly enhance radiologists’ reading performance, highlighting the potential of AI as a valuable tool in medical imaging.
The study involved a cohort of 30 doctors, including 20 board-certified radiologists and 10 radiology residents. They assessed 120 chest radiographs, with half depicting lung cancer cases and the other half showing no abnormalities. The participants were divided into two groups and analyzed the images without AI assistance in the first session. In the subsequent session, each group reinterpreted the images with the aid of either a high-accuracy or low-accuracy AI model.
Lunit INSIGHT CXR, a commercially available AI solution for chest X-ray analysis, served as the high-accuracy AI model. The low-accuracy AI model was trained using a limited dataset compared to Lunit INSIGHT CXR. The study revealed that the high-accuracy AI model significantly improved radiologists’ performance, with the AUROC increasing from 0.77 to 0.82 when assisted by the AI model.
On the other hand, the group that utilized the low-accuracy AI model did not experience any performance improvement, as the AUROC remained at 0.75. Additionally, the study found that radiologists who utilized the high-accuracy AI model were more receptive to AI suggestions, accepting 67% of AI recommendations that contradicted their initial readings compared to 59% acceptance in the low-accuracy AI group.
Importantly, the study emphasized that factors such as individual expertise, experience with AI, or attitudes toward AI had minimal influence on the radiologists’ reading performance. Instead, the accuracy of the AI model and the radiologists’ initial diagnostic accuracy were the primary determinants of the final diagnostic determination.
These findings highlight the significance of using high-performance AI models as a second reader in medical imaging. The study demonstrates how AI assistance can improve radiologists’ diagnostic accuracy and increase their acceptance of AI within medical practices. Lunit’s commitment to advancing the field of cancer diagnostics through cutting-edge technology is exemplified by their development of AI-powered solutions that augment the expertise of healthcare professionals.
With the publication of this study in ‘Radiology,’ a prestigious journal in the field of medical imaging, Lunit’s findings contribute to the growing body of evidence supporting the integration of AI into medical practices. The study underscores the potential of AI to revolutionize medical diagnostics and improve patient outcomes.