Expert-Driven Survival Machines: Improving Stratification and Interpretability in Multiple Clinical Cohorts

2026-06-12

AI Summary

Current deep survival models often employ a uniform feature representation, which struggles to capture the heterogeneity among patient subgroups, thereby limiting the accuracy and interpretability of risk stratification. To address this limitation, this work proposes the Adaptive Deep Clustering Survival Model (AdaCSM), which innovatively integrates a Mixture-of-Experts (MoE) architecture with a dynamic routing mechanism to enable patient-specific modeling within a parametric survival analysis framework. AdaCSM jointly optimizes survival prediction and subgroup clustering objectives, simultaneously producing risk estimates and uncovering clinically meaningful subtypes. Extensive experiments on multiple real-world longitudinal clinical cohorts demonstrate that AdaCSM significantly outperforms state-of-the-art methods, achieving superior predictive performance while generating individualized risk stratifications that are clinically interpretable.

Abstract

Survival prediction plays a central role for healthcare providers and clinical researchers. Accurate risk stratification enables early intervention and improved patient management. Most existing deep survival models learn one common feature representation for all patients, which may hide important differences between patient subgroups. In contrast, a Mixture-of-Experts (MoE) framework allows different parts of the model to focus on different patient patterns, leading to more individualized representations. Therefore, in this work, we propose a mixture-of-experts enhanced adaptive deep clustering survival framework (AdaCSM) for modeling such heterogeneous survival patterns. We introduce a routing-based expert mechanism that enables conditional specialization within a parametric survival modeling framework. The proposed architecture allocates patients to specialized risk predictors dynamically while preserving the patient survival and subtype clustering objectives. We compare our method with state-of-the-art survival and deep clustering models on multiple real-world longitudinal clinical cohorts spanning diverse disease domains. The proposed method demonstrates improved predictive performance and leads to interpretable results in survival analysis.

Problem

• survival prediction
• risk stratification
• patient heterogeneity
• Mixture-of-Experts
• interpretability

Innovation

• Mixture-of-Experts
• Survival Prediction
• Deep Clustering
• Risk Stratification
• Interpretability