Epilepsy EEG signals classification based on sparse principal component logistic regression model.

Area of Science:

• Neurology
• Biomedical Engineering
• Machine Learning

Background:

• Epilepsy is a chronic neurological disorder characterized by recurrent seizures due to excessive brain neuron discharge.
• Predicting epilepsy stages is vital for managing patient care and improving quality of life.
• Electroencephalogram (EEG) signals are commonly used for epilepsy diagnosis and analysis.

Purpose of the Study:

• To develop an efficient method for predicting epilepsy stages using EEG signal features.
• To address challenges of high dimensionality and multi-collinearity in EEG-derived features.
• To improve the accuracy and reliability of epilepsy stage prediction models.

Main Methods:

• Extracted Pearson correlation coefficients (PCC) between EEG channels in different frequency bands as features.
• Employed Principal Component Analysis (PCA) for dimension reduction.
• Utilized logistic regression with L1 and L2 regularization to avoid overfitting and achieve feature sparsity.

Main Results:

• The proposed method demonstrated high performance on the CHB-MIT dataset.
• Achieved an average accuracy of 94.86%, average precision of 96.71%, and average recall of 93.48%.
• Obtained an average kappa value of 0.90 and an average Matthews correlation coefficient (MCC) of 0.90.

Conclusions:

• The combination of PCA and regularized logistic regression effectively predicts epilepsy stages.
• The proposed approach offers a robust and efficient solution for epilepsy stage identification.
• This method holds significant potential for clinical application in epilepsy management.