Akademik Veri Yönetim Sistemi
Signal, Image and Video Processing, cilt.17, sa.4, ss.1255-1263, 2023 (SCI-Expanded)
Migraine is a neurological disease defined by recurrent attacks of headache accompanied by nausea and vomiting, which causes autonomic nervous system disturbance, episodes of severe pain, are prolonged, and can have a major impact on quality of life. In this study, a deep learning model was proposed to assist expert judgment in the automatic detection of migraine using electroencephalography (EEG) signals. The dataset consists of EEG signals recorded from 21 healthy and 18 migraine volunteers. Feature vectors were created by calculating the power densities values of the frequencies between 1 and 49 Hz of the EEG signals using the Welch method. The performances of bidirectional long-short term memory (BiLSTM) deep learning algorithm and random forest, support vector machine, and linear discriminant analysis machine learning algorithms in EEG-based migraine classification tasks were compared using the created feature vectors. The algorithm with the highest performance is the BiLSTM (95.99%) deep learning algorithm using 128 channels. The study is a rare attempt in which a deep learning model is used in the effective diagnosis of automatic migraine by analyzing multi-channel EEG signals and provides evidence for the superiority of deep learning algorithms. In comparison to state-of-the-art investigations, more accuracy was achieved. With this model, the accuracy was increased by 6.3%.