Epilepsy disorder detection and diagnosis using empirical mode decomposition and deep learning architecture

Abstract

Epilepsy neurological disorder is detected and diagnosed in this article using deep learning method by differentiating the focal electroencephalogram (EEG) signals and the non-focal EEG signals. The proposed method consists of time-scale signal decomposition, feature extraction, and classification with diagnosis. The empirical mode decomposition (EMD) method is used to decompose the EEG signal into six intrinsic mode function (IMF) sub bands. The external intrinsic features are computed for each of the decomposed IMF sub bands and these features along with the coefficients of each IMF sub bands are then trained and further classified using the proposed convolutional neural networks (CNN) architecture. The CNN classifier classifies the EEG signal into either focal or non-focal and finally the focal signal is diagnosed as mild case or severe case using the proposed CNN architecture. The proposed framework achieves a Se of 99.8%, Sp of 99.9%, and Acc of 99.8% on the EEG signals which are available in Bern-Barcelona dataset. The proposed framework for diagnosis of severe case signals achieves 99.2% of diagnosis rate (DR) and 99.6% of DR for mild case signals on Bern-Barcelona dataset. The proposed framework achieves a Se of 99.6%, Sp of 99.6%, and Acc of 99.7% on the EEG signals which are available in CHB-MIT dataset. The proposed framework for diagnosis of severe case signals achieves a DR of 97.3% and DR of 98.0% for mild case signals on CHB-MIT dataset. The performance of the proposed method is cross validated by k-fold method.

Epileptic Seizure Detection Using a Hybrid 1D CNN-Machine Learning Approach from EEG Data

“…They implemented a hybrid model of CNN and the long short-term memory (LSTM) network. In [ 16 ], Srinath et al used EMD to decompose signal into six IMFs. The intrinsic features were computed from these sub-bands.…”