Horizontal visibility network topology reveals pathological loss of complexity in heart rate variability

Abstract Physiological health is typically associated with complex, multiscale heart rate variability (HRV) arising from the coordinated action of regulatory mechanisms. In contrast, pathology is often marked by the progressive loss of this organization. Here we apply horizontal visibility graphs (HVGs) to RR-interval ECG time series from cohorts with Normal autonomic function, Early cardiac autonomic neuropathy (CAN) and Definite CAN. The HVG transformation represents the signal as a network in which the nodes (amplitudes) can form links regardless of their temporal distance provided a certain inequality is satisfied. We analyze total and scale-resolved triangular motifs together with the associated Euler characteristic as indicators of network organization. These measures display a systematic monotonic reduction with increasing pathology that is robust against variation in window length and in the threshold used to define long-range structure, consistent with the interpretation that pathology is associated with a simplification in graph topology. The strongest separation observed is between Normal and Definite CAN, while Early-stage separation remains more modest. Furthermore, the network measures are strongly correlated with conventional time-domain HRV, particularly RMSSD. Consequently while the present analysis does not establish independent or predictive information beyond standard HRV indices, it demonstrates that clinically associated differences in HRV are also expressed in the topology of the corresponding visibility graphs.