Skip to content

WILSON - a pathology foundation model framework for patient-level analysis and diagnostic text generation

Sep 2026 · 0 citations
Biology Computer Science Engineering

TL;DR

WILSON, a vision--language foundation model that represents whole-slide images and multi-slide cases as single multi-magnification composite images as single multi-magnification composite images, trained on approximately 189k slides from Mayo Clinic spanning 42 organs and 829 diagnostic entities using pathology reports as supervision.

Abstract

Pathologists integrate morphology across magnifications and across the slides of a patient case, whereas pathology foundation models encode thousands of tiles from single slides and aggregate their features. Here we present WILSON, a vision--language foundation model that represents whole-slide images and multi-slide cases as single multi-magnification composite images, trained on approximately 189k slides from Mayo Clinic spanning 42 organs and 829 diagnostic entities using pathology reports as supervision. Without task-specific training, WILSON exceeded a dedicated case-level model on all internal cohorts (macro-F1 0.52 versus 0.38) and matched slide-level models up to 9.4 times larger at 272- to 2,155-fold lower compute. End-to-end fine-tuning on 508 triple-negative breast cancer cases improved histologic subtyping and stromal tumor-infiltrating lymphocyte grading by 0.16 and 0.11 macro-F1. WILSON retrieved matching diagnostic text at 75.6% recall@1 (PRISM, 58.1%) and generated captions closer to report-derived references than PRISM and PRISM2 on the internal cohort and on most external comparisons. Composite images thus offer a compact, clinically aligned computational unit for pathology.

View source

Similar papers

Preprint Oct 2026

Pathology foundation models for training-free case retrieval in salivary gland tumors

This approach makes departmental archives searchable with pathology foundation models, providing a low-barrier route to case-based differential diagnosis and extending the individual frame of reference to the collective experience of the department.

Maximilian J. Roesel, N. Kiermeyer, Philipp Keyl et al. · 0 citations
Preprint Aug 2026

An end-to-end-trained vision-language model for native-language prostate pathology report generation

This work presents a slide-level framework generating prostate biopsy reports that is language-independent by construction: tokenizer and model are trained from scratch, and institutions can thus train native-language reporting models on their own archives.

Christian Grashei, Fabian Gülhan, Maximilian Legnar et al. · 0 citations
Open access Sep 2026

Segment anything in pathology images with natural language

A foundation model that uses natural language descriptions to segment structures across anatomical regions and spatial scales is presented, establishing a unified framework for flexible pathology segmentation with potential utility for clinically interpretable image analysis.

Zhixuan Chen, Jun Hou, Li-Qi Lin et al. · 1 citation
Sep 2026

SlideChat is a multimodal generative artificial intelligence assistant for whole-slide computational pathology across cancer types

SlideChat, a multimodal generative artificial intelligence assistant for whole-slide computational pathology across cancer types, outperformed leading baselines in closed-ended accuracy and report-generation Metric for Evaluation of Translation with Explicit Ordering score and received the highest expert ratings across...

Ying Chen, Chenglong Ma, Qiong-Qiong Li et al. · 2 citations
Open access Sep 2026

An evaluation of pathology foundation models with weakly supervised learning for BRAF mutation prediction in melanoma whole slide images

Overall, foundation models combined with weakly supervised learning provide competitive performance for BRAF status prediction in melanoma WSIs, but the findings highlight the need for diverse melanoma datasets and the importance of evaluating model performance beyond a single metric when developing H&E-based computati...

Lucy Godson, Jack Breen, Alexander I. Wright et al. · 0 citations
#artificial intelligence Preprint Oct 2026

Morphology-aware deep learning predicts colorectal cancer survival from whole-slide images

Background: Five-year survival prediction from hematoxylin-and-eosin whole-slide images could support risk stratification, but the task is weakly supervised, and the tissue morphologies carrying prognostic information remain incompletely characterized. Methods: We conducted a retrospective secondary analysis using TCGA...

Hikmat Khan, M. Gurcan, A. Parwani et al. · 0 citations

Related blog posts

MIT News · Artificial Intelligence Oct 28, 2026

Making machines that can listen

Professor Paris Smaragdis develops computers with superhuman hearing for applications like speech recognition, music understanding, and environmental monitoring.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.