A Mixture-of-Experts Framework with Log-Logistic Components for Survival Analysis on Histopathology Images

2025-11-09

AI Summary

This paper addresses cancer-specific survival prediction from whole-slide imaging (WSI) data, proposing a modular mixture-of-experts framework to model tumor spatial heterogeneity and complex survival distributions. Methodologically, it introduces: (1) a quantile-gated patch selection mechanism for adaptive identification of survival-relevant regions; (2) graph-guided pathological region clustering to strengthen local–global structural modeling; and (3) a hierarchical contextual attention network jointly optimized with an expert-driven mixture of log-logistic distributions for robust survival probability estimation. Evaluated on TCGA-LUAD, KIRC, and BRCA WSI cohorts, the method achieves C-indices of 0.644, 0.751, and 0.752, respectively—significantly outperforming state-of-the-art approaches. The framework advances interpretability and predictive accuracy by explicitly integrating spatial pathology priors with flexible survival distribution modeling.

Technology Category

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

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs 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: Explainable and interpretable methods for personalization

Abstract

We propose a modular framework for predicting cancer specific survival from whole slide pathology images (WSIs). The method integrates four components: (i) Quantile Gated Patch Selection via quantile based thresholding to isolate prognostically informative tissue regions; (ii) Graph Guided Clustering using a k nearest neighbor graph to capture phenotype level heterogeneity through spatial and morphological coherence; (iii) Hierarchical Context Attention to learn intra and inter cluster interactions; and (iv) an Expert Driven Mixture of Log logistics framework to estimate complex survival distributions using Log logistics distributions. The model attains a concordance index of 0.644 on TCGA LUAD, 0.751 on TCGA KIRC, and 0.752 on TCGA BRCA respectively, outperforming existing state of the art approaches.

Problem

• Predicting cancer survival from whole slide pathology images
• Identifying prognostically informative tissue regions in histopathology
• Modeling complex survival distributions using log-logistic mixtures

Innovation

• Quantile Gated Patch Selection isolates prognostic tissue regions
• Graph Guided Clustering captures phenotype heterogeneity spatially
• Mixture of Log-logistics estimates complex survival distributions