Bayesian heart-rate entropy identifies autonomic dynamics during seizure evolution: a reproducible pilot study

Epileptic seizures are accompanied by significant changes in autonomic control, yet the cardiac dynamics data remains underexplored. Bayesian heart-rate entropy, a probabilistic method to measure cardiac dynamics, has been suggested as a tool to capture clinically relevant seizure aspects, such as awareness or seizure progression.

Researchers applied the BayesianAtHeart framework to analyze electrocardiographic recordings during video-EEG monitoring. Bayesian entropy values from these recordings were combined with clinical data to create a seizure-level dataset for further analysis. The study included 51 out of 67 seizures from 10 patients, as the rest had insufficient ECG quality for analysis.

There was no link found between Bayesian HR entropy and ictal awareness when evaluating seizure averages, mixed-effects models, or time-resolved trajectories. Instead, seizure length turned out to be the main clinical factor affecting Bayesian HR entropy, with longer seizures showing a downward trend in entropy. Time-resolved analysis revealed this trend was a gradual decline in entropy throughout the seizure, rather than a low initial value.

Post hoc sensitivity checks confirmed this association was not due to selection bias or the number of intervals used in the entropy calculation. Longer seizure duration, rather than inherent patient differences, was the primary contributor to variability in entropy. The study found that entropy was not inherently lower during seizures compared to a pre-ictal state; rather, the entropy trajectory's variability diminished progressively with seizure duration.

The results suggest that Bayesian HR entropy mainly reflects the changing autonomic regulation organization during seizures, not behavioral awareness. The research also presents a fully reproducible method for Bayesian HR entropy analysis, providing a basis for future investigations into autonomic dynamics in epilepsy.