HAPI-EP: Towards Hybrid, Adaptive, and Predictive Digital Twins of Cardiac Electrophysiology

2026-06-14 · Source: Machine Learning · Field: Health & Wellbeing — Medical Devices & Health Technology, Health & Medical Research, Artificial Intelligence & Machine Learning · Depth: Expert, quick

Summary

HAPI-EP is an AI framework designed to create hybrid, adaptive, and predictive digital twins (DTs) for patient-specific cardiac electrophysiology. It tackles the core challenges of rapid DT adaptation to live data and robust predictive capability, which arise from limitations in both mechanistic and data-driven model formulations and optimization strategies that often lead to un-identifiable models. HAPI-EP achieves this through three key enablers: first, a physics-integrated gray-box model that augments an interpretable mechanistic backbone with a neural component to model residuals; second, rapid on-the-fly adaptation to few-shot live data via feedforward meta-learners for amortized inference of model parameters; and third, the construction of a conditional generative model, ensuring theoretical identifiability and strong predictive performance. The framework's proof-of-concept is demonstrated in cardiac electrophysiology using a hybrid monodomain model with mechanistic reaction kinetics and neural graph diffusion, showing enhanced predictive and out-of-distribution capabilities across synthetic and real-data studies.

Key takeaway

For AI scientists developing patient-specific digital twins, HAPI-EP's approach offers a robust solution to adaptation and prediction challenges. You should consider integrating physics-informed gray-box models with meta-learning for rapid, few-shot adaptation to live data. This method enhances model identifiability and predictive power, crucial for clinical applications. Explore its hybrid monodomain model for cardiac electrophysiology to improve your DT's out-of-distribution capabilities.

Key insights

HAPI-EP builds identifiable, predictive digital twins by integrating mechanistic and neural models with meta-learning for rapid adaptation.

Principles

• Hybrid models improve interpretability and accuracy.
• Meta-learning enables rapid DT adaptation.
• Predictive objectives enhance model identifiability.

Method

HAPI constructs a physics-integrated gray-box model, then uses feedforward meta-learners for amortized inference of parameters, trained with predictive objectives to adapt to few-shot live data, forming a conditional generative model.

In practice

• Apply gray-box models in biophysics.
• Use meta-learners for patient-specific DTs.
• Validate DTs with out-of-distribution data.

Topics

• Digital Twins
• Cardiac Electrophysiology
• Hybrid Models
• Meta-learning
• Physics-informed AI
• Generative Models

Best for: AI Scientist, Research Scientist

Related on AIssential

• Graphical conditional generative modeling for digital twin modeling — A framework discovers essential variables for parsimonious stochastic surrogates by analyzing full conditional laws, not just means.
• KairosHope: A Next-Generation Time-Series Foundation Model for Specialized Classification via Dual-Memory Architecture — KairosHope is a TSFM using dual-memory and statistical features for specialized time-series classification.
• Learning Cardiac Electrophysiology Digital Twins Through Agentic Discovery of Hybrid Structure — An LLM agent discovers stable, personalized cardiac electrophysiology digital twin models using structured domain knowledge.
• Echo2ECG: Enhancing ECG Representations with Cardiac Morphology from Multi-View Echos — Aligning global ECGs with multi-view Echos resolves representational mismatch, yielding powerful, lightweight cardiac morphology features.
• HAFMat: Hybrid Priors Guided Adaptive Fusion for Single-Image Human Material Estimation — HAFMat uses hybrid priors and adaptive fusion to overcome the ill-posed problem of single-image human PBR material estimation.
• Evolving Systems, Volume 17, Issue 1, February 2026 — The volume showcases diverse applications of AI and optimization algorithms across engineering, medical, and agricultural domains.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.