On the viability of ECG features for seizure anticipation on long-term data

Besides the evident brain state alterations present in electroencephalogram (EEG), epileptic seizures are also associated with changes in the cardiovascular status. In particular, heart rate (HR) has become an important autonomic biomarker in seizure prediction. Based on that, a preliminary study is here proposed in order to inspect the behaviour of electrocardiogram (ECG) derived features in the period preceding epileptic seizures. The study took place using data from 1275 seizures collected from a set of 167 patients available in EPILEPSIAE database. The analysis was conducted considering three different variables: seizure type, seizure hour onset and vigilance state. The results did not reveal a clear effect of any of the three variables, assessed individually, in entire seizure set. Nevertheless, some evidence has been found that, for some seizures, it was possible to detected a consistent pattern of increase/decrease in feature magnitude before the onset. These patterns were revealed using the mean of RR intervals and the mean of the number of beats per minute.