Bipartite Patient-Modality Graph Learning with Event-Conditional Modelling of Censoring for Cancer Survival Prediction

2025-07-22

AI Summary

To address insufficient utilization of censored data and unstable performance under multimodal missingness in cancer survival prediction, this paper proposes CenSurv. Methodologically, it (1) constructs a patient–modality bipartite graph to explicitly model cross-modal associations; (2) introduces an Event-Conditioned Censoring Modeling (ECMC) module that dynamically identifies high-confidence censored samples and reconstructs survival times via reweighting; and (3) incorporates modality alignment and dynamic momentum accumulation to enhance robustness to missing modalities. Evaluated on five public datasets, CenSurv achieves an average 3.1% improvement in concordance index (C-index) over state-of-the-art methods. Moreover, integrating the ECMC module into eight baseline models yields an average 1.3% C-index gain, demonstrating significant improvements in stability and generalization under multimodal missingness.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Reasoning under Uncertainty: Graphical Models

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Accurately predicting the survival of cancer patients is crucial for personalized treatment. However, existing studies focus solely on the relationships between samples with known survival risks, without fully leveraging the value of censored samples. Furthermore, these studies may suffer performance degradation in modality-missing scenarios and even struggle during the inference process. In this study, we propose a bipartite patient-modality graph learning with event-conditional modelling of censoring for cancer survival prediction (CenSurv). Specifically, we first use graph structure to model multimodal data and obtain representation. Then, to alleviate performance degradation in modality-missing scenarios, we design a bipartite graph to simulate the patient-modality relationship in various modality-missing scenarios and leverage a complete-incomplete alignment strategy to explore modality-agnostic features. Finally, we design a plug-and-play event-conditional modeling of censoring (ECMC) that selects reliable censored data using dynamic momentum accumulation confidences, assigns more accurate survival times to these censored data, and incorporates them as uncensored data into training. Comprehensive evaluations on 5 publicly cancer datasets showcase the superiority of CenSurv over the best state-of-the-art by 3.1% in terms of the mean C-index, while also exhibiting excellent robustness under various modality-missing scenarios. In addition, using the plug-and-play ECMC module, the mean C-index of 8 baselines increased by 1.3% across 5 datasets. Code of CenSurv is available at https://github.com/yuehailin/CenSurv.

Problem

• Predict cancer survival accurately using multimodal data and censored samples
• Address performance degradation in modality-missing scenarios during inference
• Improve robustness and accuracy via event-conditional censoring modeling

Innovation

• Bipartite graph models patient-modality relationships
• Event-conditional modeling handles censored data
• Complete-incomplete alignment for missing modalities