Gastric Cancer Pathological Image Segmentation based on Convolutional Neural Network

¹ School of Electrical and Information Engineering, Jiangsu University of Technology, Changzhou Jiangsu, China
² Department of Pathology, Changzhou First People's Hospital, Changzhou, Jiangsu, China

^a zhangwenyue97wren@163.com
^* Corresponding author: jiaziyan@jsut.edu.cn
^b liqblk@163.com
^c zhangdachuan@suda.edu.cn
^d jsjshedy@jsut.edu.cn
^e sdw@jsut.edu.cn

Abstract

The difficulty in pathological image diagnosis of gastric cancer lies in the accurate segmentation of cancerous tissue in the picture. To increase gastric cancer pathological images segmentation accuracy, we optimized the basic UNet model and proposed the DCU-Net model. First, add a direct channel module to each layer of the encoding part to obtain more detailed information. In addition, considering that the image may cause a loss of information during the transmission process, a CA module is added before the up-sampling and down-sampling of each layer so that the model can obtain more channel information. After conducting segmentation experiments on our own gastric cancer data set and comparing it with several current classic segmentation models. The experiment proved that the segmentation accuracy using this article's model can be enhanced on our own gastric cancer image segmentation data set, achieving an accuracy of 91.30% and an IoU of 79.88%.