Patient-dependent epilepsy seizure detection using random forest classification over one-dimension transformed EEG data

This work presents a computational method for improving seizure detection for epilepsy diagnosis. Epilepsy isthe second most common neurological disease impacting between 40 and 50 million of patients in the world and it proper diagnosis using electroencephalographic signals implies a long and expensive process which involves medical specialists. The proposed system is a patient-dependent offline system which performs an automatic detection of seizures in brainwaves applying a random forest classifier. Features are extracted using one-dimension reduced information from a spectro-temporal transformation of the biosignals which pass through an envelope detector. The performance of this method reached 97.12% of specificity, 99.29% of sensitivity, and a 0.77 h −1 false positive rate. Thus, the method hereby proposed has great potential for diagnosis support in clinical environments.