Surface and intracranial EEG spike detection based on discrete wavelet decomposition and random forest classification

Epilepsy is a neurological disorder for which the electroencephalogram (EEG) is the most important diagnostic tool. In particular, this diagnosis heavily depends on the detection of interictal (between seizures) paroxysmal epileptic discharges (IPED) in the EEG. This is a time-consuming task requiring significant training and experience. Automatic detection of these EEG patterns would greatly assist visual inspections of human readers. We present a new method, which allows automatic detection of IPED based on discrete wavelet decomposition and a random forest classifier. The algorithm was trained and cross validated using 17 subjects with scalp EEG and 10 subjects with intracranial EEG. The performance of this method reached 62% recall and 26% precision for surface EEG subjects and 63% recall and 53% precision for intracranial EEG subjects. Thus, the method hereby proposed has great potential for diagnosis support in clinical environments.