Gastric Pathology Image Recognition Based on Deep Residual Networks

Abstract

Gastric cancer is a malignant neoplasm with a high mortality rate in the world. Nearly one million new cases occur each year. The most important measure to diagnose gastric cancer is the detection and treatment of diseases early. Gastric cancer detection is currently performed by pathologists reviewing large expanses of biological tissues, but this process is labor intensive and error-prone. In this paper, a framework for automatically detection of tumors in gastric pathology image (slide) has been proposed based on deep learning. A deep residual network with 50 layers is built by identity mapping on a dataset of pathology images. The proposed method makes the training of models easier and improves the generalization performance. Finally, the experimental results show that the F-score of our method achieves 96%. The research in auto-classification of gastric pathology images has great value for gastric cancer detection in clinical medicine.