Automatic onset detection of abnormal ventricular potentials by time-frequency analysis

Orrù, Marco;Mandas, Nicla

;Baldazzi, Giulia;Viola, Graziana;Pani, Danilo

2025-01-01

---

Abstract

Activation times in intracardiac electrograms (EGMs) may help clinicians in detecting abnormal pathways sustaining post-ischemic ventricular tachycardia (VT). As they are especially accurate when estimated only from the near-field component of abnormal ventricular potentials (AVPs), this work proposes a novel approach for AVP onset identification, leveraging the Hilbert-Huang Transform (HHT) and a knee-point detection technique. A dataset composed of 940 delineated AVPs was used. The HHT was exploited to compute an expedient signal as the product between the instantaneous frequency and energy, only when the frequency fell within a specific range. Values ranging from 0 to 490 Hz were tested for the lower limit, while the upper limit was fixed to 500 Hz. On the expedient signal, the cumulative area is computed, and the AVP onset was identified as its knee-point. Based on the absolute errors distributions and the percentage of errors lower than 10 ms, the optimal lower frequency threshold resulted in 110 Hz, with a median error in the onset delineation of 13.5 ms. This approach could be relevant for electrophysiological studies aimed at VT suppression, by enabling the identification of near-field components in AVPs, and a more accurate assessment of their local activation timing.