Artificial Neural Network analysis of EEG waves in complex partial seizure patients

Background: Brain dynamics associated with epilepsy remains limited. EEG-based epilepsy diagnosis and seizure detection is still in its infancy. The problem is further amplified for the design and development of automated algorithms, which requires a quantitative parametric representation of the qualitative or visual aspect of the markers. This study proposes an automatic classification system for epilepsy based on neural networks and EEG signals. Material and Method: The present study made use of EEG data from 16 controls and 16 temporal lobe epilepsy (TLE) patients in order to comparatively assess neural dynamics in normal healthy young adults and epileptic patients treated with anti-epileptic drugs in the context of resting state during eye close session. Such tangible differences could be appreciated through artificial neural network (ANN) classifiers. Results: During eye closed session of EEG in order to diagnose temporal lobe epileptic patient, the extracted features of EEG activity are given to the classifier algorithm for training and test performance. Artificial Neural Network (ANN) classifier was used for the diagnosis task. Fractal dimension (Katz, Higuchi and Permission entropy) were analyzed, in which the best results was observed in trained set of data of Katz (93.18%).   Conclusion:  Non-linear analysis plays an important role in prediction of complex partial seizure during interictal period.