Survival Modeling from Whole Slide Images via Patch-Level Graph Clustering and Mixture Density Experts

2025-07-22

AI Summary

This study addresses cancer-specific survival prediction from whole-slide imaging (WSI) data. We propose a modular deep learning framework that jointly models histopathological phenotypic heterogeneity and complex, multimodal survival distributions. Methodologically, it integrates dynamic thresholding for patch selection, graph-guided k-means clustering, hierarchical attention mechanisms, and a Mixture of Density Experts (MoDE) model—replacing conventional single-distribution assumptions with explicit modeling of survival time uncertainty. On TCGA-KIRC and TCGA-LUAD, our method achieves C-indices of 0.712 and 0.645, respectively, significantly outperforming state-of-the-art approaches. Our key contributions are: (i) the first integration of graph-based clustering and MoDE for WSI-level survival analysis; (ii) improved generalizability across cancer types; and (iii) enhanced clinical interpretability through biologically grounded clustering and uncertainty-aware survival estimation.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Mixture of Experts (MoE)
• Reasoning under Uncertainty: Graphical Models

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

We introduce a modular framework for predicting cancer-specific survival from whole slide pathology images (WSIs) that significantly improves upon the state-of-the-art accuracy. Our method integrating four key components. Firstly, to tackle large size of WSIs, we use dynamic patch selection via quantile-based thresholding for isolating prognostically informative tissue regions. Secondly, we use graph-guided k-means clustering to capture phenotype-level heterogeneity through spatial and morphological coherence. Thirdly, we use attention mechanisms that model both intra- and inter-cluster relationships to contextualize local features within global spatial relations between various types of tissue compartments. Finally, we use an expert-guided mixture density modeling for estimating complex survival distributions using Gaussian mixture models. The proposed model achieves a concordance index of $0.712 pm 0.028$ and Brier score of $0.254 pm 0.018$ on TCGA-KIRC (renal cancer), and a concordance index of $0.645 pm 0.017$ and Brier score of $0.281 pm 0.031$ on TCGA-LUAD (lung adenocarcinoma). These results are significantly better than the state-of-art and demonstrate predictive potential of the proposed method across diverse cancer types.

Problem

• Predict cancer survival from whole slide pathology images
• Improve accuracy via patch selection and graph clustering
• Model survival distributions using Gaussian mixture models

Innovation

• Dynamic patch selection via quantile-based thresholding
• Graph-guided k-means clustering for phenotype heterogeneity
• Attention mechanisms for intra- and inter-cluster relationships