Annu Int Conf IEEE Eng Med Biol Soc. 2025 Jul;2025:1-5. doi: 10.1109/EMBC58623.2025.11253910.
ABSTRACT
Detecting abnormal electrical wave propagation in Sinus Rhythm (SR) in patient with Ventricular Tachycardia (VT) can enhance treatment strategies. However, traditional methods, like activation maps, often produce biased results due to the difficulty of calculating what is called Local Activation Time (LAT). This paper introduces a method based on Network Granger Causality (NGC) to assess VT mechanics. Instead of computing LATs, NGC generalizes the binary Granger causality framework and allows to leverage the EGM morphologies, identifying temporal causality links within a network of electrograms. Furthermore, by incorporating geometrical information in the NGC model, we improved the interpretation of causal relationships and applied our method to localize anomalies via the reconstruction error computation. Using simulated clinical data, we compare our method to standard techniques, such as voltage and velocity computation. Our results show that NGC-based analysis improves VT anomaly detection during SR over standard methods in terms of ROCAUC, increasing performance from 0.80-0.85 to 0.89. Finally, we present an example on a real patient with VT.
PMID:41337213 | DOI:10.1109/EMBC58623.2025.11253910