A Novel Spectral Density Operator Approach to Unveil Dynamic Time Dependencies in Multivariate Long-Term ECGs

Abstract

Electrocardiograms (ECGs)are essential tools for assessing cardiac health. The extraction of meaningful features from multivariate ECG data is critical for identifying potential cardiac anomalies based on interchannel dynamics. This work introduces an innovative approach rooted in functional data analysis (FDA) aimed at analyzing the temporal and frequency-domain dynamics of multivariate ECG signals. Considering each patient’s heartbeats as a sample observation of a multivariate functional time series, this method employs the spectral density operator to reveal dynamic time-dependencies among ECG channels across frequency bands. Through the analysis of ECG data of a subject experiencing detachment from the self, this approach demonstrates its potential to uncover clinically relevant patterns that correlate with the patient’s symptomatology.