Mitosis Detection in Breast Cancer Using Deep Learning
Mary Jane C. Samonte, Qian Bian, Zhenduo Wang, Chen Xiao-fan · 2023
Breast cancer is a common gynecological malignant tumor. At the same time, breast cancer is also a cancer that can be detected early and prevented early. Using efficient mitotic cell detection technology for breast cancer screening and diagnosis is significant. The application of computer-aided analysis technology to cell detection tasks in pathological medical images has gradually become a hot research direction. In this study, the object detection technique based on YOLO v5 was first applied to the field of mitotic cell detection in pathological breast images. Aiming at the three difficult problems cell detection faces in pathological breast images, this paper explores the influence of detection models on domain generalization from the perspective of image preprocessing. A series of experimental studies show that the color correction strategy based on CycleGAN adopted in this study applied to the YOLO v5 detection model can effectively improve the model's applicability on the external test set.