Scattering Transform of Heart Rate Variability for the Prediction of Ischemic Stroke in Patients with Atrial Fibrillation

Abstract

Atrial fibrillation is an identified risk factor for ischemic strokes. Thus, the dynamics of heart rate under fibrillation might provide a useful predictor for strokes. The complex, nonlinear and multiscale nature of the heart rate calls for the use of powerful signal processing techniques. We explored the application of a novel tool, the scattering transform, to discriminate and predict ischemic strokes from heart rate dynamics. We found that groups of scattering coefficients, at several time scales, were significantly higher (p-value < 0.05) in patients who developed ischemic strokes than in those who did not. We also found significant differences in predictive power (C-statistic) between scattering coefficients and the CHA2 DS2-VASc score. Results suggest the use of scattering coefficients for the analysis of patients with atrial fibrillation.