Modeling the dynamics of the heart rate variability by hidden Markov models

A study of the nonlinear dynamics of the heart rate variability (HRV) was done using hidden Markov models (HMM). The HRV was obtained from 24-hours Holter-ECG recordings. The RR series were selected from 6-hour night period of patients with idiopatic dilated cardiomiopathy (IDC) and healthy subjects (NRM). Two groups of patients were considered in the IDC group: HR, patients with high risk of developing sudden cardiac death (SCD); LR, patients without SCD. In the present study, HMMs were identified from the words generated applying symbolic dynamics to the RR series. An alphabet of 4 symbols was considered and words of 3 symbols were constructed from the transformed RR series to symbols. Different HMM topologies were analyzed. The logarithm of the observation sequence probability given the model and the maximum probability of the distinct observed words in each state could characterize the HR and LR groups.