Majority Vote for Electroencephalography (EEG)-Based Migraine Classification

Migraine (MD) is a neurological disorder that can present with auditory and visual symptoms known as aura, affecting approximately one billion people globally. This condition causes temporary disability and can progress to serious diseases such as epilepsy or stroke, resulting in significant losses in society's productivity. The overlap of migraine symptoms with other illnesses complicates the diagnosis process for medical professionals. To improve healthcare and patient care beyond traditional methods, we have developed a machine learning model to help doctors diagnose and differentiate between migraine types, with and without neurological aura. The model uses EEG signals from visual stimuli and analyzes them using discrete wavelet transform (DWT) to extract frequency bands: alpha, beta, delta, theta and gamma. The data is then augmented without exceeding the original frequency bands. Each participant's data is organized into a matrix, with rows representing channels and columns for the frequency bands. A majority voting mechanism determines the final classification; if most channels indicate a specific type of neural activity, the participant is classified accordingly. Our model achieved a classification accuracy of 90.58%, effectively diagnosing migraines and distinguishing their main types. By integrating advanced signal processing with machine learning, our model represents a significant advancement in migraine diagnosis and enhances patient care.