Receiver-operating-characteristic analysis reveals superiority of scale-dependent wavelet and spectral measures for assessing cardiac dysfunction

Receiver-operating-characteristic analysis was used to assess the suitabili ty of various heart-rate variability (HRV) measures for correctly classifyi ng electrocardiogram records of varying lengths as normal or revealing the presence of heart failure. Scale-dependent HRV measures were found to be su bstantially superior to scale-independent measures (scaling exponents) for discriminating the two classes of data over a broad range of record lengths (hours to minutes). A jittered integrate-and-fire model built around a fra ctal Gaussian-noise kernel provides a realistic, though not perfect, simula tion of heartbeat sequences.