Nuclei Segmentation and Count in Breast Pathology Image with Deep Learning

Abstract

Nuclei quantitative analysis is significance in pathology image diagnosis such as Hematoxylin and eosin staining (H&E) slides and immunohistochemistry (IHC). In recent years, deep learning, in particular convolutional neural networks (CNN), have rapidly become a methodology of choice for nuclei segmentation. Training model requires datasets of images in which a vast number of nuclei have been annotated, however, there are few annotated IHC images. Here, we proposed a CNN model trained by publicly accessed annotated H&E images, and IHC slides can be tested after color deconvolution and normalization with color distribution of H&E image. Model showed better performance in nuclei segmentation than open source software with testing breast cancer H&E and IHC images from hospital. This result suggests the application potential of our new method for nuclei quantitative analysis, which is objective and efficient.