Migraine disease diagnosis from EEG signals using Non-linear Feature Extraction Technique

Migraine is a prolonged neurovascular illness, which causes outbreaks of severe pain and autonomic nervous system disturbance. The clinical analysis of Electroencephalogram signals helps in management and prognosis of migraine disease. Recent advancement in biomedical signal processing field led to generation of various techniques for multi-resolution analysis of Electroencephalogram signals and diagnosis of diseased condition. In present work, a nonlinear parametric approach of Electroencephalogram feature extraction is proposed and analysed for automated diagnosis of migraine disease. The Electroencephalogram database studied in present study was prepared in SMS Hospital, Jaipur, India. The database contains Electroencephalogram activity record of 26 healthy and migraineurs subjects. The Permutation Entropy, Higuchi's Fractal Dimension and Katz Fractal Diemension based features are extracted from processed Electroencephalogram signals. The extracted Electroencephalogram activity is classified using SVM, ANN and RF classifiers. It is illustrated from the classification results that the classification accuracy of 88% is achieved in migraine disease diagnosis task in present work.