Locating ictal activities over human scalp with automated detection using EEG signals

摘要

Epilepsy is one of the most common brain disorder which affects more than sixty-five million people worldwide. Its diagnosis is usually performed using electroencephalography technique. Since, electroencephalogram (EEG) signals are highly susceptible to artifacts, epilepsy diagnosis is often challenging task. In addition, locating the origin of seizure patterns is even more tedious due to the non-linear nature of EEG signals. Therefore, automated seizure detection systems are highly important to overcome the said challenges. In this study, we have designed a procedure which integrates the seizure detection and localization process with ceiling level of efficiency. Here, we evaluated the energy and standard features using `coiflets' wavelet packets. Then, the feature matrix was reduced by applying fast correlation based filter. Application of 5-Nearest Neighbour classifier resulted in mean accuracy of 99.3515±0.2518 % taking only 0.152±0.0344 seconds for execution. Later, independent component analysis over the ictal segments was applied and subsequently topographic scalps maps were plotted. The results successfully implied automated detection and visualization of ictal activities over the human scalp.