Journal of Digital Information Management Siyue HAN, ZhizhaoYU, Jiahui YU, Simon James Fong, Yongze GUO, HongweiZHANG, Huicong DONG, Saicong LU, Xian WANG, Feng TIAN 24 1 2026 https://doi.org/10.6025/jdim/2026/24/1/1-10 https://www.dline.info/fpaper/jdim/v24i1/jdimv24i1_1.pdf Gastric cancer is among the leading causes of cancer related mortality worldwide, and its early detection remains challenging due to subtle endoscopic manifestations. Artificial intelligence (AI) assisted analysis of endoscopic images offers a promising pathway for improving diagnostic accuracy in real time. This paper presents a real time feature extraction framework for fine mucosal structures (FMS) in gastroscopy images using a structure enhanced morphological watershed segmentation algorithm. The method performs image preprocessing through directional filtering and morphological operations to highlight FMS patterns, followed by marker controlled watershed segmentation to deline ate structural boundaries. Geometric features including area, circularity, convexity, and Feret diameter are quantified and evaluated using multiple machine learning classifiers. Experimental results on endoscopic datasets demonstrate that filtering segmentation outliers significantly improves classification performance, with neural networks achieving an AUC of 0.813. The results indicate that max Feret diameter is the most discriminative feature for distinguishing early gastric cancer from non cancerous tissues. This study contributes to real time intelligent healthcare systems by providing an interpretable and computationally efficient feature extraction pipeline that can support AI-driven clinical decision making in endoscopic screening.