Back

Provenance-Aware Explainable Digital Twin for Personalized Health Management

Mohsin, M. T.; Abdulrashid, I.

2025-12-04 health informatics
10.64898/2025.12.03.25339981 medRxiv
Show abstract

AI can now help with personalized prediction, tracking, and decision-making thanks to progress in data-driven health analytics. However, many models are still hard to understand in clinical settings. To address these limitations, our work presents the Provenance-Aware and Explainable Digital Twin (PA-XDT) framework, integrating a digital twin, explainable AI techniques, and transparent provenance tracking for patient-centered health management.PA-XDT uses a compact LSTM-based twin trained on short temporal sequences to model near-term physiological dynamics and quantify uncertainty. In the implemented system, this twin works alongside a Gradient Boosting classifier that provides stable risk predictions, supported by global and local SHAP analyses and twin-validated counterfactual checks. A lightweight provenance layer records hashed inputs, outputs, and explanation metadata, enabling verifiable audit trails. Experiments on a gallstone risk dataset show that this combined pipeline improves physiological coherence and maintains strong predictive performance.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.