A Comparison of Segmentation Methods in Gastric Histopathology Images.

Gastric Cancer is one of the five most common types of malignant tumors among men and women worldwide and it is very important to make precise diagnosis for the early stage of gastric cancer. In this paper, we compare eight methods in Gastric Histopathology Image Segmentation (GHIS) including most classical and state-of-the-art ones. For estimating the segmentation result, we use seven evaluation indexes. Our study carries out that deep learning method shows the effectiveness in GHIS and the DenseCRF using the U-Net feature map performs best overall.