AI-BAC helped doctors find more breast lesions on mammograms

What happened

Deepwise built AI-BAC, software whose deep learning module detects suspicious calcifications, masses, asymmetries and architectural distortions on mammograms of women over 40. The Center for Medical Device Evaluation (CMDE) of China's National Medical Products Administration reviewed the company's application for Class III registration. The trial was the maker's own, a randomized crossover trial at three institutions with 855 cases enrolled. Doctors read each case with and without the software, and three breast imaging experts set the reference by consensus. Lesion-level sensitivity was 87.32% with the software and 66.20% without it. Subject-level specificity was 73.51% with the software and 70.70% without it, which met the non-inferiority test. On 3 January 2023, CMDE concluded that “该产品临床使用受益大于风险” (the product's clinical benefits outweigh its risks) and recommended approval. As long-term safety and effectiveness still needed follow-up, it recommended that Deepwise keep studying how well the algorithm generalizes and analyze the product's use every year. The reference was the experts' reading, and the report mentions no pathology.