Research of Deep Learning on Gastric Cancer Diagnosis

Abstract

Gastroscopy is the first method to check gastrointestinal cancer and related diseases. Traditional manual methods have the disadvantages of time-consuming, high missed diagnosis and misdiagnosis rates. Image recognition technology based on deep learning has great potential in improving diagnosis efficiency and accuracy. We summarize the latest advances in deep learning in gastric cancer diagnosis from the aspects of data sets, image preprocessing, and classification algorithms. Most research teams cooperate with hospitals to build small data sets, and preprocess the images using threshold-based filters and other methods. In terms of gastric cancer and gastric disease classification, the DenseNet model has the highest ACC, F1 score and specific SP.