Predicción de la fase pre-ictal de convulsiones en pacientes con epilepsia a partir de señales electroencefalográficas y electrocardiográficas

Abstract

Seizures are harmful to patients, who, without timely prediction, can lead to death. Therefore, having algorithms that indicate when an epileptic episode is going to occur provides security and action time to act. The present work focuses on the prediction of seizures in patients with epilepsy from electroencephalography (EEG) and Electrocardiography (ECG) signals. The study was carried out in patients who suffered seizures and Machine Learning algorithms were implemented for the prediction of the pre-ictal phase of seizures using the "Class Learner" tool from Matlab. For the development of the work, the CRISP-DM methodology was used, with which characteristics of 10 patients can be extracted in order to train different classification algorithms. The EEG and EKG signals were considered together and separately to show which of the two obtained better performance according to the metrics computed from the confusion matrix. It was shown that the best sensitivity was obtained when the characteristics extracted from the EEG and EKG were worked together.