Local Activation Identification in Persistent Atrial Fibrillation Intracardiac EGM Signals for Automatic Spatio-temporal Dispersion Pattern Recognition.

Atrial fibrillation (AF) is a common cardiac condition that predominantly affects the elderly population, presenting a significant risk factor for strokes and thus raising concerns in public health. Catheter ablation (CA) stands out as the most effective long-term treatment for persistent AF. A recently proposed novel CA approach is based on spatio-temporal dispersion (STD). This technique targets the STD patterns associated with active zones responsible for sustaining the arrhythmia. In this work we want to solve the peak detection problem, since it is a fundamental step for the automatic classification of STD patterns from multipolar electrograms (EGM). Instead of using machine learning models which lacks explainability, we want to understand the classification process performed at the block by interventional cardiologists in real time. The scenario is very challenging because the STD classification relies on visual peak detection to identify local activations, which are used to measure if STD does occur or not. We present our peak detector comparing it with eight different techniques from the state of the art. To extract peaks from real intracardiac EGM signals is difficult, most classical signal processing methods fail. We evaluate a total of nine techniques on the challenging scenario of real STD data. We analyze if the peaks are correctly identified, being part of the mathematical pipeline. Results show that identifying the peaks is a fundamental aspect to built the presented mathematical pipeline to overcome the STD classification problem, improving the classification accuracy with respect to previous works.