Impact of Artificial Intelligence on Accuracy and Reading Time in Breast Mammogram Interpretation

Abstract Background Breast cancer is the second leading cause of death from cancer in women Bray, but early detection and treatment can considerably improve outcomes. As a consequence, many developed nations have implemented large-scale mammography screening programmes. Aim of the Work to evaluate the impact of artificial intelligence on reading time and accuracy in breast mammogram interpretation. Patients and Methods This study is carried as retrospective comparative study conducted at the radiology department, Ain Shams University Hospitals, the main source of data for this study were the patients referred to the radiology department at Ain Shams university hospitals for mammography from august 2022 to October 2023. Results the most experienced radiologist had the highest accuracy, sensitivity and specificity (97.5%, 95% and 100% respectively). The next 2 radiologists had comparatively lower overall accuracy (92.5% and 92.5% for radiologist 2 and 3), with the more experienced of them showing higher specificity and lower sensitivity figures than the less experienced (100% Versus 95% for specificity and 85% Versus 90% for sensitivity). On the other hand, the AI was inferior to the most experienced radiologist for sensitivity and overall accuracy figures and showed equal specificity whereas it showed non inferior figures to the next 2 radiologists demonstrating relatively higher sensitivity and overall accuracy figures compared to both radiologists. Conclusion We found that the AI accuracy has been proven to be comparable to the radiologist 1 (highest years of experience) with equal specificity and lower sensitivity, and not inferior and even higher than radiologist 2 and 3 (medium and low years of experience). The reading time of the AI software was much less than the three radiologists. The highest agreement of the AI software was with the radiologist 1 (the most experienced).