Seizure prediction using statistical dispersion measures of intracranial EEG

Researches indicate that electrophysiological changes develop minutes to hours before the actual onset of epileptic seizures due to abnormal neuronal discharges. These precursors perceived through symptoms like sleep problems or headaches are observable from the analysis of the intracranial electroencephalogram (iEEG). It can be utilized as a major tool for seizure prediction well in advance. In this work an algorithm with a statistical feature set consisting of mean absolute deviation (MAD) and inter quartile range (IQR) is proposed to predict epileptic seizures. A linear classifier has been used to find the seizure prediction time in preictal iEEGs. A sensitivity of 100% with zero false positive rate (FPR) in 12 patients and very low values of FPR for the rest were achieved using widely used Freiburg iEEG dataset. Average prediction time varies between 51 and 96 min.

► Mean absolute deviation and inter quartile range were used as discriminating features for predicting epileptic seizures. ► Prediction was treated as a classification between preictal and interictal EEGs, which was done using linear classifier. ► The method was tested with widely used Freiburg University EEG dataset. ► The method showed 100% sensitivity. False positive rate (FPR) was zero in 12 cases and very low in rest of the cases. ► Comparison with other works on the same database showed improvement in terms of sensitivity, prediction time and FPR.