Accepted for/Published in: JMIR Medical Informatics
Date Submitted: Sep 7, 2025
Date Accepted: Aug 13, 2026
Warning: This is an author submission that is not peer-reviewed or edited.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 Reinforcement Learning-based Temporal Knowledge Graph Reasoning Model for Chronic Gastritis Diagnosis Prediction
ABSTRACT
Background:
The clinical progression of chronic gastritis entails intricate temporal dependencies, making it difficult for conventional methods to capture both the disease’s dynamic trajectory and the underlying relationships among medical events.
Objective:
To predict chronic gastritis diagnosis dynamically, we proposed RL4TKGR, a Reinforcement Learning-based Temporal Knowledge Graph Reasoning model for chronic gastritis diagnosis prediction.
Methods:
Specifically, RL4TKGR incorporated an innovative reinforcement learning policy network that employed a dual-path encoding module to separately model historical and non-historical diagnostic information. Furthermore, RL4TKGR integrated a dual-channel reward function with a dynamic weight allocation mechanism, which adaptively balanced the two information sources, thereby addressing the strategic bias problem and enabling interpretable reasoning.
Results:
Experiments conducted on the self-constructed Chronic Gastritis Temporal Knowledge Graphs (CG-TKG) demonstrated that RL4TKGR delivered state-of-the-art performance across MRR and Hits@1/3/10 metrics, outperforming existing baseline models. Ablation experiments and case analyses further validated the efficacy of each module design and the model's practical utility in predicting disease subtypes and therapeutic medications.
Conclusions:
This paper advanced the predictive accuracy of chronic gastritis while establishing a novel and effective paradigm for interpretable reasoning over complex temporal medical data.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.