AICOR: Comprehensive Database of Atrial Fibrillation and Regular Atrial Rhythm Simulations for AI-Driven Cardiac Electrophysiology

This dataset originates from the ai.cor project (Industrial Research Project for new tools for epicardial potential estimation and treatment guidance through AI in patients with cardiac arrhythmia – Ref: 2021/C005/00146416), funded by Red.es (Ministry for Digital Transformation and Public Function, Government of Spain) and the European Union (NextGenerationEU, within the framework of the Recovery, Transformation and Resilience Plan). The primary objective of the ai.cor project was to develop and validate novel Artificial Intelligence (AI) tools to enhance the precision of cardiac mapping devices and improve the efficacy of arrhythmia treatments. A core component of this project was the creation of this extensive database, essential for the training and validation of the ai.cor.ecgi and ai.cor.treat AI modules. This database comprises a massive and diverse collection of simulated "digital twins" of cardiac electrophysiology, specifically focusing on atrial fibrillation (AF) and regular atrial rhythms. It was meticulously generated through advanced mathematical modeling techniques, allowing for the simulation of various patterns of electrical propagation and different ablation strategies that would be impossible to replicate in real-world patients. Key features of this dataset include: Scale and Diversity: Over 138,000 in-silico "digital twins" were generated, significantly exceeding the initial project target of 50,000. This large volume and diversity ensure robust training and validation for deep learning models. Realistic Cardiac Models: The dataset is built upon realistic mathematical models of atria and torso geometries, capable of simulating various electro-anatomical and structural variations (e.g., fibrotic tissue, altered ionic properties, reduced intercellular coupling). Comprehensive Electrophysiological Scenarios: Baseline Simulations: Includes simulations of normal sinus rhythms, ectopic beats, and complex arrhythmias like atrial fibrillation. Ablation Strategy Simulations: Crucially, the dataset contains over 103,428 ablation simulations, encompassing 12 distinct anatomical and functional ablation strategies (e.g., Pulmonary Vein Isolation - PVI, BOX ablation, rotor-guided ablation). Each simulation is meticulously labeled with its outcome (success/failure). This provides invaluable ground truth for training predictive AI models like ai.cor.treat. Variability for Robustness: Data augmentation techniques (rotations, scaling, geometrical transpositions, addition of Gaussian white noise to BSPM signals) were systematically applied to enhance the robustness and generalization capabilities of the trained AI models against clinical variability and noise. Data Structure: The dataset includes body surface potentials (BSPMs) and corresponding epicardial electrograms (EGMs), providing a comprehensive input-output relationship for inverse problem solutions. Availability for Research Use: In alignment with the "broad dissemination of results" principle ("Amplia Difusión") of the Red.es funding program, Corify Care is committed to fostering open science and collaboration. This dataset is made publicly available for non-commercial research purposes only. Researchers from academic institutions and non-profit organizations are encouraged to utilize this valuable resource to advance the understanding of cardiac electrophysiology, develop new AI algorithms, and contribute to the scientific community. For any inquiries regarding this dataset or potential collaborations, please contact [Insert a general research contact email for Corify Care, e.g., [e-mail] or [e-mail]]. Funding: This work was supported by the Ministry for Digital Transformation and Public Function, Government of Spain, through Red.es (Project Ref: 2021/C005/00146416), and co-funded by the European Union through the NextGenerationEU instrument, within the framework of the Recovery, Transformation and Resilience Plan. DOI: 10.5281/zenodo.15807037