Complexity measurements for analysis and diagnosis of early ventricular fibrillation

We present complexity measurements from ECG signals during ventricular fibrillation (VF) using the Gaussian kernel algorithm. Ten ECG data sets of early VF among 53 pig subjects are selected for such analysis. A single test uses a segment of 33.33 seconds and 15 measurements are made on each ECG trace over sliding windows with a skipping length of 10 seconds. It has been shown that the early VF contains ~80%-90% low-dimensional deterministic dynamics and ~10%-20% high-dimensional component. Dimensions differ from subject to subject and segment to segment, with D¯₂¯~5, and entropy is positive. For a given ECG trace, variation, in dimension along the data is within ±1.0 around its mean value and no monotonic trend exists. Further, the measurements reveal the existence of complex structure within 3 minutes, which is consistent with periodicity analysis