Machine learning-based image analysis for accelerating the diagnosis of complicated preneoplastic and neoplastic ductal lesions in breast biopsy tissues

Diagnosis of breast preneoplastic and neoplastic lesions is difficult due to their similar morphology in biopsy specimens. To diagnose these lesions, pathologists perform immunohistochemical analysis consult with expert pathologists. These additional examinations are time-consuming expensive. Artificial intelligence (AI)-based image has recently improved, may help ordinal pathological diagnosis. Here, we showed the significance machine learning-based for facilitating high-throughput Images were obtained from normal mammary glands, hyperplastic such as usual ductal hyperplasia (UDH), columnar cell lesion (CCL), carcinoma situ (DCIS), DCIS comedo necrosis (comedo DCIS) The original enhanced convoluted neural network (CNN) system was used analyzing images. AI-based provided following area under curve values (AUC): versus DCIS, 0.9902; 0.9942; CCL, 0.9786; UDH 1.000. Multiple comparison precision recall scores those single analysis. Based on gradient-weighted class activation mapping (Grad-CAM) visualize important regions reflecting result CNN analysis, ratio stromal tissue whole weighted significantly higher CCL than that DCIS. analyses provide a more accurate rapid diagnosis patients. Moreover, Grad-CAM identifies uncharted histological characteristics newer findings targets research understanding diseases.