Detection of ictal patterns in electroencephalogram signals using 3D phase trajectories

Epilepsy is one of the most common brain disorder. Its diagnosis is generally performed using electroencephalography (EEG). But, the process of visually inspecting the EEG signals to trance out the seizure patterns has always been very time consuming and difficult. Researchers have continuously tried to automate the process of recognizing ictal patterns. But, the complexity and high erroneous outcomes from most of the deployed expert systems continues to limit practical realization. In this study, authors have presented a simple yet novel methodology to detect ictal patterns in EEG signals. The proposed model used intrinsic mode functions to plot 3D phase trajectories. The mean values of Euclidean distances calculated from these trajectories were used as input feature vectors to probabilistic neural network classifier. The outcomes from the classifier discriminated between the seizure and seizure-free patterns with ~98 % accuracy. The model achieved similar statistical performance in <; 0.03 s. Thus, attesting the design for application in practical settings.