Abstract

Info

Title:: Gastric tumor segmentation algorithm based on global-local attention mechanism

Author(s):: XU Kaicheng1; FANG Zhijun1; CAI Qingping2; WEI Ziran2; GAO Yongbin1; JIANG Xiaoyan1; 1. School of Electronic and Electrical Engineering, Shanghai University of Engineering Science, Shanghai 201600, China 2. Shanghai Changzheng Hospital, Shanghai 200003, China

Keywords:: gastric tumor upper abdominal CT deep learning medical image segmentation GLat-Net

Abstract:: Using CT to realize preoperative gastric tumor diagnosis is a potentially efficient technical method, of which accurate tumor image segmentation is the prerequisite for preoperative diagnosis. In order to achieve a better performance on the extraction of tumor, a 2D segmentation network (GLat-Net) based on attention mechanism is proposed to complete the segmentation of the gastric tumor on upper abdominal CT images, which focuses on the area around the tumor and extracts effective contextual information from global and local perspectives. Furthermore, the weight module is introduced to decoder module for highlighting the representative features. The experiment results show that the proposed approach outperforms current state-of-the-art methods on gastric tumor segmentation, having a higher segmentation accuracy.