CT Image Prediction Of PD-1 Gastric Cancer Patients Based On The PLSG Framework

Abstract

Survival prediction in PD-1 inhibitor patients has received extensive attention in recent years. Existing diffusion models generally focus blurring on key lesion regions, and the masks are weakly matched to CT images during the sampling process, resulting in a relatively high risk of misdiagnosis. For this reason, a new framework for late survival prediction of gastric cancer (PLSG) is designed in this paper. A mask-based adaptive diffusion algorithm is proposed for enhancing the characterization of critical regions in medical images. A three-domain feature mapping module based on a self-attention mechanism is designed to efficiently acquire higher-order features of images. A sparsification-based principle is utilized to select high-quality features in the prediction stage, which in turn achieves the final screening and prediction of features. Numerous experiments show that the model has greater value in PD-1 treatment.