Modified wavelet bicoherence as a diagnostic tool for very high frequency peaks in cardiovascular signals of normal and heart transplant subjects

Wavelet bicoherence (WB) provides insight into nonlinear interactions and can provide information about interacting oscillations in the cardiovascular system, especially in the very high frequency (VHF) band. We suggest a new normalization for the WB estimator which enables a combined threshold criteria reducing false positive error from 5% to below 1%. Alternatively, we suggest a modified wavelet bicoherence (MWB) based on frequency shifts using Hilbert transform and a mixer which enables an optimal time resolution. Simulations were performed to assess the threshold values and to support the usability of the proposed methods. Implementation on heart rate signals of heart transplant (HT) and normal subjects indicated improved sensitivity as compared to Fourier based bicoherence. Significant peaks were found in the VHF band in 60% of the HT group and, for the first time, in normal subjects, in 40% of the control group.