Programa del congreso

Idiopathic ventricular arrhythmias (IVA) are commonly treated with radiofrequency ablation, where procedural success relies on accurate identification of the effective ablation point (EAP). In current clinical practice, EAP selection is largely based on the cardiologist’s expertise in interpreting intracardiac electrograms (EGMs), making the process operator-dependent and difficult to standardize. Developing automatic, data-driven approaches to characterize and localize the EAP could improve reproducibility and assist clinical decision-making.

In this work, we present an extended analysis of EGM dynamics near ablation points in IVA patients. Building on our earlier pilot study of three cases, we now analyze a curated database of 48 procedures in which the EAP was prospectively annotated. For each case, we identified regions of interest around the annotated EAP, preselected points in close spatial proximity, and compared their electrogram properties to those of distant points. To characterize temporal and morphological signal features, we extracted descriptors from the TSFEL library and applied statistical analysis to identify features discriminative of the EAP. In parallel, we explored nonlinear signal representations by constructing latent spaces using Multiple Kernel Learning (MKL), with dynamic time warping employed as a kernel metric to capture temporal variability in EGMs.

This study represents one of the largest systematic efforts to date to investigate EGM-based signatures of effective ablation sites in IVA. While ongoing, our work highlights the potential of combining comprehensive time-series feature extraction with kernel-based representations to advance the development of automated, standardized tools for guiding ablation therapy.