A review and comparison of breast tumor cell nuclei segmentation performances using deep convolutional neural networks.

摘要: Breast cancer is currently the second most common cause of cancer-related death in women. Presently, clinical benchmark diagnosis tissue biopsy examination. However, manual process histopathological analysis laborious, time-consuming, and limited by quality specimen experience pathologist. This study's objective was to determine if deep convolutional neural networks can be trained, with transfer learning, on a set images independent breast segment tumor nuclei breast. Various were evaluated for study, including U-Net, Mask R-CNN, novel network (GB U-Net). The trained Hematoxylin Eosin (H&E)-stained eight diverse types tissues. GB U-Net demonstrated superior performance segmenting sites invasive diseases (AJI = 0.53, mAP = 0.39 & AJI = 0.54, mAP = 0.38), validated two hold-out datasets exclusively containing approximately 7,582 annotated cells. results networks, tissue, that could accurately segmented.