Real Time Power Spectral Analysis of Heart Rate Variability

ECG monitoring of heart rate variability (HRV) in a noninvasive method for the evaluation of cardiovascular autonomic control. HRV can be analyzed in the time- and frequency domain. The frequency-domain analysis (FDA), an estimation of the variability distributed as a function of frequency, is performed with spectral techniques using the Fast Fourier transform (FFT) or autoregressive (AR) algorithms. With FDA three main rhythmic components: the VLF (0-0.03 Hz), LF (0.04-0.15 Hz) and HF (0.15-0.40 Hz). HF measures the amplitude of sinus node respiratory arrhythmia (vagal modulation); LF mainly reflects the effectes of sympathetic discharge. Physiological interpretation of VLF is still uncertain. FDA is applied to short term (5 min.) and long term (24 hours) recordings. In both cases FDA requires some level of stationarity of the signals during the time window, usually chosen for evaluation. This impedes the quantification of clinically relevant transient variations of sympathovagal balance. Different time-frequency methods have been developed for instant FDA of HRV. Those based on short-time FFT still require some level of stationarity. In this work an autoregressive (AR) adaptive parametric method, based on the Yule-Walker equation, has been used, which is suitable for real-time FDA during data acquisition. ECG signal (band-passed 0.05-100Hz) is digitized with a 500 Hz sampling frequency. The RR interval is computed from QRS first derivative with threshold method. The AR adaptive software program automatically the optimal order of the model and tests its reliability. The LF and HF central frequency, the total power, the LF and HF power in absolute values and in normalized units and the LF/HF ratio are computed beat-to-beat. The results of beat-to-beat spectral analysis is typically represented three-dimensionally. The power spectrum at each beat can be separately analyzed with automatic calculation of all parameters. The method has been applied to investigate normal volunteers and patients with syncope, Parkinson Disease and Multisystem Atrophy. All patients have been studied in supine position, under spontaneous and metronome-controlled breathing, during Valsalva maneuver, and in active 90\ub0 upright position. The average modifications of spectral pattern of HRV were the same as those previously reported with conventional FFT or AR algorithm. The real time study of the RR interval dynamics has identified transient changes of HRV spectral components, which might helpful for clinical investigation of neural mediation of syncope, dysauthonomia and of several cardiac arrhythmia