Accepted for/Published in: JMIR Medical Informatics
Date Submitted: Sep 8, 2025
Date Accepted: Feb 9, 2026
(closed for review but you can still tweet)
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
A Bi-layer Feature Fusion Framework for Pan-cancer Survival Prediction Based on Multi-head Attention and Adaptive Differential Privacy: Model Development and Validation Study
ABSTRACT
Background:
In the field of precision medicine, pan-cancer survival prediction is crucial for individualized oncology diagnosis and treatment. Although multimodal data fusion techniques have significantly improved prediction accuracy, existing studies generally overlook the sensitivity of medical data and the need for privacy protection.
Objective:
To address the aforementioned problem, this study aims to propose a bi-layer feature fusion framework based on the multi-head attention mechanism and adaptive differential privacy, which balances precise feature extraction and sensitive data protection.
Methods:
Specifically, the multi-head attention mechanism was integrated into bi-layer feature extraction and fusion. Layer-wise relevance analysis was used to calculate the correlation between features and outcomes, and Laplacian noise was adaptively added based on the calculation results to achieve collaborative optimization of precise feature extraction and sensitive data protection. Additionally, the concordance index and 5-fold cross-validation were employed to compare the proposed method with state-of-the-art approaches.
Results:
The proposed model achieved superior performance in both pan-cancer and single-cancer survival prediction, validated via the concordance index (C-index) and 5-fold cross-validation. In pan-cancer scenarios, the tri-modal combination of clinical, mRNA, and miRNA data yielded the highest C-index of 0.799, surpassing state-of-the-art methods such as MultiSurv (0.771) and Ziling Fan (2023) (0.777). After adaptive Laplacian noise injection for privacy protection, the model’s accuracy decreased by only 0.01–0.03 while satisfying ϵ-differential privacy. For single-cancer prediction, compared with existing deep learning methods (Cheerla and Gevaert, 2019), the model achieved higher C-index values in 18 out of 20 cancer types, with 7 types showing significant improvements (C-index difference > 0.1) and more stable performance distributions. Furthermore, pan-cancer-trained models generally outperformed single-cancer-trained counterparts in most cancer types, highlighting the value of shared predictive features across cancers.
Conclusions:
This study provides a solution that balances prediction accuracy and privacy security for pan-cancer survival prediction, laying the foundation for the efficient utilization of medical data under privacy protection. Future work may further integrate pathological images and proteomics data to expand the model’s applications in cancer subtype classification and biomarker discovery.
Citation
