A Convolutional Neural Network Combined with Color Deconvolution for Mitosis Detection.

Mengjie Bai,Jiarui Sun,Kaili Cheng,Jingting Luo,Yong Zhao
DOI: https://doi.org/10.1145/3297156.3297168
2018-01-01
Abstract:Since deep learning has achieved an epochal success in many vision task, many researchers have proposed applying deep learning to some pathological images in digital pathology. The mitosis detection is an important but tedious task. We propose an approach for automatic mitosis detection from pathological images. First, we use a stain normalization procedure on image data. Then, we select image patch in pathological images. We use color deconvolution to get the color space features based on the stain, and then combine it with original RGB image to train in the improved ResNet. The approach fuses traditional color features into deep neural networks. Experiments in TUPAC 2016 MITOSIS datasets show that our method achieves a great performance.
What problem does this paper attempt to address?