ViT-based AI automates pancreatic tumor microenvironment analysis; LSR predicts survival

MacrinePhD · x · 2026-09-16

In npj Digital Medicine (2026), Xiawei Li, Yulian Wu and colleagues built a multicenter Vision Transformer model on whole-slide images to automate tissue segmentation and quantify the lymphocyte-stroma ratio (LSR) in pancreatic ductal adenocarcinoma. Higher LSR correlated significantly with improved overall survival, while low-LSR groups faced markedly higher risk (multicenter P < 0.001). The team also released an accessible online platform for automated WSI segmentation, replacing a manual, subjective workflow.

Original post →

More from Research

Research channel →