Dear friends and members of the ESC WG on e-Cardiology,

Welcome back to all interested readers to the fourth issue of our selection on relevant publications in the world of e-cardiology. In this issue, we focus on recommendations and fundamentals of artificial intelligence (AI) in clinical electrophysiology (EP), the opportunities and challenges of generative AI in cardiac EP medical education as well as review of AI and digital tools for design and execution of cardiovascular clinical trials.
 
The European Heart Rhythm Association (EHRA), the Heart Rhythm Society (HRS), and the ESC Working Group on e-Cardiology recently published a checklist to standardise the reporting of AI research in cardiac EP[1]. This important initiative aims to enhance transparency, reproducibility, and clarity in AI-related EP studies. The developed 29-item checklist was applied to 55 studies across three domains of atrial fibrillation management, sudden cardiac death, and EP lab applications. Interestingly, findings revealed that no domain reported more than 55% of the checklist items, with critical areas like trial registration, participant details, data handling, and training performance often underreported (<20%). In future, the EHRA AI checklist should serve as a structured framework to improve reporting practices, thereby facilitating the integration of AI into clinical EP practice. 
 
In a brief review, Hanycz and Antiperovitch address risks and challenges of generative AI to create new content to optimise trainee education in electrophysiology [2]. Exemplarily, generative adversarial networks (GANs) are presented to create realistic, synthetic surface electrocardiogram (ECG) signals that can be used for training purposes on specific diseases. The concept of GANs consists of 2 neuronal networks of which, in this example, one tries to create an artificial ECG that cannot be identified as such by a discriminator which is trained based on a labelled dataset of ECG. This approach might in future be used to create ECGs to improve training and diagnosis e.g. of life-threatening diagnosis, including hyperkalemia, long QT, and ventricular tachycardia. However, as stated by the authors, once generative AI can reliably generate ECGs, it must undergo thorough testing in medical education to assess its ability to improve learner outcomes and retention.
 
In the third selected publication, Hu et al. provide a comprehensive state-of-the-art review summarising the opportunities and challenges associated with digital tools and AI in clinical trials [3]. Digital tools are broadly defined to include a wide range of technologies used to facilitate tasks in medicine, including electronic data capture systems, innovative wearable and sensor-based technologies, and smartphone-based health applications. The potential and limitations of these tools to improve case screening, patient outreach, intervention, and data analysis are thoroughly discussed. The authors concluded that, if these digital tools are implemented successfully, they hold great promise in revolutionising the efficacy, accuracy, and cost-effectiveness of clinical trials in cardiovascular medicine. 
 
Finally, Julia Ramirez concludes our selection with her comments on an important study that developed and externally validated an AI-ECG model on a comprehensive paediatric and adult population with congenital heart disease to predict imaging-defined left ventricular systolic dysfunction [4].
 
We hope you enjoy reading the selected works. 
 
On behalf of the entire ESC WG on e-Cardiology,
 
Sven Knecht, PD DSc,
Nucleus member of the ESC WG on e-Cardiology