Improving Breast Cancer Detection in Screening Mammography with Artificial Intelligence Assistance: A Multi-reader Retrospective Study

Introduction: This study aimed to compare the performance of radiologists in screening mammography for breast cancer detection, with and without artificial intelligence (AI) assistance, including subgroup comparison between breast radiologists and general radiologists in Hong Kong.Methods: This was a single-centre multi-reader retrospective study.A screening mammography test set was used (the Hong Kong Personal Performance in Mammographic Screening Scheme), comprising 80 mammograms with negative or benign findings and 36 mammograms with pathologically proven breast cancer acquired from December 2009 to December 2023.Radiologists' performance with and without AI assistance from a commercially available tool (Lunit INSIGHT MMG) was evaluated from December 2023 to April 2024.The two reading sessions were separated by a 4-week washout period.Study endpoints included sensitivity and specificity in the mammographic detection of breast cancer.The Obuchowski-Rockette model was used to estimate and compare diagnostic accuracy.Results: A total of 16 radiologists completed the test set, including nine (56.3%) breast radiologists and seven (43.8%) general radiologists.Without AI assistance, the overall sensitivity and specificity in breast cancer detection were 73.3% and 89.9%, respectively.With AI assistance, both metrics improved significantly to 80.7% (p = 0.007) and 94.3% (p < 0.001), respectively.Subgroup analysis showed that breast radiologists demonstrated improved specificity from 87.6% to 92.6% (p < 0.001), while general radiologists acquired more sensitivity from 54.0% to 66.7% (p < 0.001) with the use of AI.Conclusion: AI assistance significantly improved the diagnostic accuracy of breast radiologists and general radiologists in screening mammography for breast cancer detection.