Deep learning for classification and localization of early gastric cancer in endoscopic images. (January 2023)

• Record Type: Journal Article
• Authors: Ma, Lingyu; Su, Xiufeng; Ma, Liyong; Gao, Xiaozhong; Sun, Mingjian
• Is Part Of: Biomedical signal processing and control. Volume 79(2023)Part 2
• Publication Date: 2023-01
• Subjects: Deep learning -- Early gastric cancer -- Convolutional neural network -- Endoscopic images -- Automatic diagnosis; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2022.104200
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Based on the prepared dataset, a complete workflow for endoscopic images EGC-related classification and segmentation was introduced. Image recognition was perfomed based on backbone networks. Simultaneously, combined with the pixel-level features of EGC, an effective guided attention module was designed to effectively extract EGC-realted feature regions. Based on widely used U-Net, a lightweight attention module and multi-scale feature extractor were introduced to assist in automatic pathology localization of EGC and evaluate the semantic segmentation performance by comparing with other models. A pilot study in endoscopic videos was conducted to verified the feasibility and practicality of the proposed computer-aided detection system. Abstract: Gastric cancer, as a malignant tumor, is one of the most common cancer-related deaths worldwide with high mortality and incidence rates. Therefore, the endoscopic detection of gastric cancer at an early stage is essential. In this paper, we propose an automatic diagnosis method of early gastric cancer (EGC) based on deep learning (DL) techniques. Specifically, with the new annotated endoscopic image dataset collected from a single-center, this paper designs several DL architectures to realize the automatic analysis of EGC images. Particularly, a guided-attention deep network was introduced, derived from ResNet-50, for the accurate score prediction of EGC and the extraction of feature information. Furthermore, we combined a lightweight attention module and multi-scale feature extractor with U-Net for estimating the pixel-level segmentation of EGC pathological regions. Experiment results on the presented dataset showed outstanding performance in the involved classification and segmentation assignments with an accuracy of 98.84% and an intersection over union (IOU) of 0.64, revealing the potential applications of DL in aiding and improving EGC diagnosis using endoscopic images.