2021 Volume 30 Issue 2 Pages 121-130
Conventional CAD(computer-aided detection / diagnosis)has been commercialized mainly in mammography and has achieved some results, but there are not a few problems that hinder its widespread use such as performance. Deep learning is a new technology that overcomes this hurdle, and CAD is now undergoing further evolution / diversification as AI-CAD. Recent prominent journals include those that show that AI-CAD performs as well as or better than physicians, those that compare the performance of multiple AI-CADs and verify their effectiveness, those that show that AI-CAD can reduce the number of false positives and the number of recalls compared to conventional CAD, and those that show the effectiveness of AI-CAD in automatically classifying normal and possibly abnormal images in screening(triage type) to significantly reduce the workload of physicians. The CEO of ScreenPoint Medical in the Netherlands predict that the performance of mammography AI-CAD currently on the market will reach the average radiologist's interpretation level, and will reach the level of high-level radiologists or higher around 2025. However, even so, the current situation is that the performance will be further improved compared to “AI-CAD” alone or “doctor” alone, not by automatic diagnosis, but by the collaboration of “AI-CAD + doctor”. This paper describes the progress of the latest AI-CAD, the problems to be solved, and the future utilization of the new AI-CAD in breast cancer screening, based on the contents of the author's keynote lecture, with some of the contents updated.
