A gastric precancerous lesion high-risk population screening method fusing clustering and machine learning

Abstract

The present application relates to the technical field of medical detection and artificial intelligence, and particularly relates to a gastric precancerous lesion high-risk population screening method fusing clustering and machine learning. The method integrates serum markers, helicobacter pylori typing, endoscopic images and clinical baseline and other multi-dimensional time sequence data, constructs a time sequence data cube through fusion processing; generates individual risk evolution trajectories based on a mixed model of a Transformer and a CGAN and a group pathologic prior constraint; after identifying typical evolution modes through unsupervised clustering, constructs a dedicated early warning model for each mode, realizes accurate determination of high-risk populations and mapping of individualized intervention suggestions, and improves the accuracy and clinical practicability of screening.

Images