SEIZURE PREDICTION BASED ON MORPHOLOGICAL FEATURES USING ONE-DIMENSIONAL LOCAL BINARY PATTERN ON LONG-TERM EEG SIGNALS

Abstract

After stroke, epilepsy has become the most prevalent disease in the world. Recurrent seizures are the hallmark of epilepsy, which is an aberrant neural activity of a brains. This illness causes abnormal, uncontrollable brain and body activity that can put a person in a coma or possi bly cause death. Electroencephalography (EEG) is frequently used to diagnose epilepsy. By analysing the EEG data to find features that will detect the aberrant activity, seizures can be detected. This study analyses EEG signals to find the greatest qualities that can foretell seizures far before they start. Early seizure recognition can reduce the severity of injuries and help in the treatment of epilepsy patients. There are several ongoing studies in this field, but the ac curacy of the results is very low in all of them. Therefore, increase the performance the study, LBP is being utilised to anticipate epileptic seizures by extracting morphological features. LBP assigns a sample point a perfect decimal value by weighting the binary values after quantizing the adjacent sample with current sample point. These LBP values help to capture the EEG signal’s rising and falling edges, using the sum of absolute difference of the LBP values and the interquartile range processed EEG signals per epoch is determined. K-nearest neighbour clas sifier is utilised for classification, and performance is assessed using data from the CHB-MIT continuous EEG dataset. In the subject wise experiment an average best accuracies of 93.66%, and in the subject dependent experiment, the best accuracies of 92.22%, and in the subject independent experiment, the best accuracies of 81.16% is obtained form KNN classifier