Get Started
Home
Topics
Search
Library
Research questionHow can pathology foundation models predict MSI-H across cancer types despite organ-specific tissue architecture?Pathology foundation models can perform well within a cancer cohort yet transfer poorly when tissue architecture and immune organization differ across organs. Scanner and site variation can further obscure features associated with MSI-H status.
AI
Computer Vision
Health
Machine Learning
Research Paper
Latest papersRecent research connected to this question, newest first.Conserved Immune Topology Improves Pathology Foundation Model Generalization for Cross-Cancer MSI-H PredictionThe study evaluates zero-shot cross-cancer and cross-site MSI-H prediction using multiple instance learning on frozen foundation-model embeddings, tile coordinates, and unsupervised immune-associated spatial descriptors, without annotations or target-domain data. Evidence comes from CPTAC-COAD and TCGA-STAD cohorts and reports improvements across three MIL aggregators, including an AUC increase from 0.6627 to 0.7161 for TransMIL.research paper · Sep 4, 2026
Related questions
How can whole-slide image classifiers preserve irregular tissue regions while modeling relationships across spatially separated regions?How can pathology foundation models be compressed to reduce inference cost without sacrificing classification accuracy or reliability?How can we identify tumor-containing slides when each patient has many whole-slide images?How can multimodal models predict lung cancer survival when imaging, clinical, and genomic data are inconsistently missing over time?