CoMetaPNS: Continually Meta-learning Personalized Neural Surrogates for Cardiac Electrophysiology Simulations
2026-06-05 • Machine Learning
Machine Learning
AI summaryⓘ
The authors address problems in creating personalized virtual heart models that are both adaptable and efficient to use. Current methods struggle when new patient data arrives over time because they must retrain on old and new data to avoid forgetting previous knowledge. The authors propose a new approach that continually learns from new data without forgetting and can recognize if the data comes from an already known or new heart behavior. They use a special statistical model to manage and identify data patterns over time, improving simulation accuracy and speed. Tests on synthetic heart data show their method works better than existing ones in handling ongoing data and avoiding forgetting.
personalized virtual heart simulationsneural surrogatesfew-shot generative modelingmeta-learningcatastrophic forgettingcontinual learningBayesian Gaussian Mixture Modelmeta-learned amortized inferencesimulation forecastingmemory buffer
Authors
Ryan Missel, Xiajun Jiang, Linwei Wang
Abstract
Personalized virtual heart simulations face challenges in model personalization and computational cost. While neural surrogates offer state-of-the-art solutions, they typically address either efficient personalization or training generalizable models. Recent work reframes this by learning the process of personalizing a surrogate using limited subject-specific context data, through few-shot generative modeling with set-conditioned surrogates and meta-learned amortized inference. These methods, however, assume a static and diverse training distribution with known task identifiers. When new data becomes available, they require costly retraining with all prior data to avoid catastrophic forgetting - a phenomena where the model forgets earlier tasks when trained on new ones. This is a major limitation in clinical settings where often unlabeled data arrives sequentially and full retraining is infeasible. This paper presents a new continual meta-learning framework to achieve personalized neural surrogates able to not only continually integrate information but also identify whether incoming data stems from a known or unknown dynamics source. By leveraging a continual Bayesian Gaussian Mixture Model over a memory buffer, our framework can infer the identifiers and relationships of data over time - required for effective meta-learning. Empirical results on synthetic cardiac data demonstrate superior simulation forecasting, computational scalability, and resilience to catastrophic forgetting compared to existing baselines.