Related Experiment Videos
Wavelet neural network classification of EEG signals by using AR model with MLE preprocessing.
Abdulhamit Subasi1, Ahmet Alkan, Etem Koklukaya
1Department of Electrical and Electronics Engineering, Kahramanmaras Sutcu Imam University, Karacasu Kampusu, 46601 Kahramanmaraş, Turkey. asubasi@ksu.edu.tr
Summary
Wavelet neural networks (WNN) offer superior accuracy for classifying epileptic seizures from EEG data compared to traditional logistic regression and feedforward artificial neural networks (ANNs). This advancement aids physicians in diagnosing epilepsy.
Area of Science:
- Biomedical Engineering
- Computational Neuroscience
- Medical Informatics
Background:
- Electroencephalography (EEG) is crucial for epilepsy therapy and diagnosis.
- Developing automated decision support systems for EEG analysis is essential.
Purpose of the Study:
- To compare the accuracy of logistic regression, feedforward error backpropagation artificial neural networks (FEBANN), and wavelet neural networks (WNN) for classifying epileptic seizures from EEG signals.
- To introduce a novel classifier architecture combining autoregressive modeling with WNN.
Main Methods:
- EEG signals were processed using Fast Fourier Transform (FFT) and autoregressive (AR) models with maximum likelihood estimation (MLE).
- Classification models including logistic regression, FEBANN, and WNN were developed and evaluated.
- Performance was assessed using Receiver Operating Characteristic (ROC) curves and scalar measures.
Main Results:
- The WNN-based classifier demonstrated superior performance over FEBANN and logistic regression.
- The novel AR with MLE and WNN architecture proved reliable for EEG signal classification.
- WNN achieved higher accuracy in distinguishing epileptic from non-epileptic seizures.
Conclusions:
- WNN-based classifiers provide a more accurate and reliable approach for automated EEG seizure detection.
- The integration of AR modeling with WNN offers a promising direction for epilepsy diagnosis support systems.