Electrogram duration for the identification of abnormal ventricular potentials

Orrù M.;Cossu M.;Baldazzi G.;Viola G.;Pani D.

2025-01-01

---

Abstract

One of the major challenges for the treatment of post-ischemic ventricular tachycardia (VT) patients undergoing radiofrequency catheter ablation is the accurate characterization of arrhythmogenic substrates. Intracardiac electrograms (EGMs) provide several quantitative metrics that can be used in decision support systems to assist clinicians during these procedures. Among these metrics, EGM duration represents a promising metric for identifying slow conduction areas, that are critical for their role in sustaining VT. By adopting an existing algorithm for the computation of EGM duration, this study aims at evaluating its effectiveness as a metric to differentiate abnormal ventricular potentials (AVPs) from physiological EGMs. The algorithm was tested on the ARGO dataset, which contains 1962 EGMs from nine post-ischemic VT patients. Statistical analyses were performed to investigate the discriminative power of EGM duration. The 70-ms cutoff proposed in the literature was evaluated on this dataset to assess its applicability. Even though the threshold demonstrated high sensitivity (98.3 percent), its specificity (14.7 percent) and overall accuracy (60.7 percent) were insufficient for reliable AVP recognition. To address this aspect, a further analysis identified an optimized cutoff of 80 ms, which slightly improved classification performance. Furthermore, EGM duration was investigated as a standalone feature for a decision-tree classification model. The model achieved an accuracy of about 72.0 percent, with balanced specificity (68.0 percent) and sensitivity (75.2 percent) in a 10-time 10-fold cross-validation scheme. These findings emphasize the potential of the EGM duration as a valuable metric for the purpose. However, in comparison to existing methodologies, this study highlights the need for multiple EGM features to enhance model robustness and clinical applicability.