Identification of inter-ictal activity in novel data by bagged prediction method using beta and gamma waves

Diagnosis of epilepsy primarily involves understanding cautious patient history and assessment of EEG (Electro Encephalography), which is an essential diagnostic support tool. It captures the electrical activity in the brain, which enables the neurologist to look for the presence of epileptiform patterns for which brain waves (Delta, Theta, Alpha, Beta, and Gamma) are studied thoroughly. The Delta (0–4 Hz), Theta (4–8 Hz), and Alpha (8- < 13 Hz) waves are interpreted visually with proficiency; however, the interpretation of Beta (13–35 Hz) and Gamma (36-44 Hz) presents a grave challenge because of their high-frequency nature. The objective of this study was to find out if these waves incorporate features essential for the identification of inter-ictal activity. The bandpass filter was used to extract beta and gamma frequency from the complete EEG signal. Five nonlinear features were extracted out from two, and four-second segments of Beta and Gamma waves. Bagged Tree Classifier is used to categorize the segments into controlled and inter-ictal activity. Data from a total of forty-two patients were used in this study; twenty-three patients with different types of epilepsy and nineteen controlled patients. For two-second segments, we achieved 91.3% classification accuracy, and for four-second segments, we achieved 93.1%. This is improvement from the previous work available in the literature where the segment length of 23.6 s has been used by researchers; with respect to use of public data. Also, the contribution of these brain waves have not been studied independently.