In this work, we used deep learning solutions to predict optimized parameters for pulsed field ablation treatments. This procedure is a new electromagnetic technique used to non-thermally ablate arrhythmogenic cardiac tissue, reducing the side effects associated with microwave thermal ablation. The precision of the lesion, induced by irreversible electroporation of the cardiac tissue, is determined by a large number of parameters (both electromagnetic and biological) that define the pulsed field ablation protocol. The purpose of our work is to predict the various parameters that define pulsed field ablation protocols in an optimized manner using artificial intelligence models. We demonstrated that deep learning methods can optimize and predict these parameters, reducing the need for computationally expensive simulations or trial-anderror approaches, which are both time-consuming and costly. The predictions were based on current literature, and the final estimated accuracy is sufficient to suggest that this approach is promising for the development of future patient-specific protocols. © 2025 European Microwave Association (EuMA).

Cited by: 0; Conference name: 55th European Microwave Conference, EuMC 2025; Conference sponsors: ASML; et al.; FNS 6G; IEEE Antennas and Propagation Society (AP-S); Keysight Technologies; The European Space Agency (ESA); Conference location: Utrecht; Conference code: 217142; Conference date: 2025-09-23 through 2025-09-25