GrapHist: Graph Self-Supervised Learning for Histopathology

2026-02-24

AI Summary

Existing self-supervised vision models struggle to effectively capture the biological characteristics of cells and their complex interactions in histopathological images. To address this limitation, this work proposes a structure-aware self-supervised learning framework that, for the first time, models tissue as a cell graph by integrating masked autoencoders with heterophilic graph neural networks to explicitly represent the heterogeneity of the tumor microenvironment. We introduce the first large-scale benchmark for self-supervised learning on pathological graphs and release five high-quality graph datasets. Experimental results demonstrate that the proposed method achieves performance comparable to state-of-the-art vision models on various downstream tasks while using 75% fewer parameters, and significantly outperforms fully supervised graph-based models.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Computer Vision: Medical and Biological Imaging
• Reasoning under Uncertainty: Graphical Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Self-supervised vision models have achieved notable success in digital pathology. However, their domain-agnostic transformer architectures are not originally designed to account for fundamental biological elements of histopathology images, namely cells and their complex interactions. In this work, we hypothesize that a biologically-informed modeling of tissues as cell graphs offers a more efficient representation learning. Thus, we introduce GrapHist, a novel graph-based self-supervised learning framework for histopathology, which learns generalizable and structurally-informed embeddings that enable diverse downstream tasks. GrapHist integrates masked autoencoders and heterophilic graph neural networks that are explicitly designed to capture the heterogeneity of tumor microenvironments. We pre-train GrapHist on a large collection of 11 million cell graphs derived from breast tissues and evaluate its transferability across in- and out-of-domain benchmarks. Our results show that GrapHist achieves competitive performance compared to its vision-based counterparts in slide-, region-, and cell-level tasks, while requiring four times fewer parameters. It also drastically outperforms fully-supervised graph models on cancer subtyping tasks. Finally, we also release five graph-based digital pathology datasets used in our study at https://huggingface.co/ogutsevda/datasets , establishing the first large-scale graph benchmark in this field. Our code is available at https://github.com/ogutsevda/graphist .

Problem

• histopathology
• cell graphs
• self-supervised learning
• tumor microenvironment
• representation learning

Innovation

• graph self-supervised learning
• cell graph
• heterophilic graph neural network
• tumor microenvironment
• histopathology representation learning