Migraine types identification based on EEG signals using machine learnings techniques

Abstract

Migraine (MD) is a neurological disorder that can be accompanied by auditory and visual symptoms called aura, affecting the lives of approximately one billion people worldwide. This condition causes temporary disability and may progress to serious diseases such as epilepsy or stroke, affecting both individual health and societal productivity by leading to a significant loss of working hours. The overlap of migraine symptoms with those of various other diseases makes identifying and diagnosing migraines challenging and time-consuming for medical professionals. To advance healthcare and improve the medical care provided to patients beyond traditional methods, which are often cumbersome and time-consuming, we developed a machine learning model to assist doctors in diagnosing migraines and distinguishing between its types, whether accompanied by neurological auras or not. The model utilizes EEG signals obtained from auditory stimuli (A) and visual stimuli (V) of 17 migraine patients and 20 healthy control (HC) subjects. These EEG signals were analyzed using discrete wavelet transform (DWT) to extract frequencies known as alpha, beta, delta, theta, and gamma. These frequency features were then used to train machine learning algorithms. Our model achieved a classification accuracy exceeding 90%, effectively diagnosing migraines and distinguishing between its main types. This innovative approach not only enhances the accuracy and efficiency of migraine diagnosis but also provides valuable insights into the neurological underpinnings of the disorder. By integrating advanced signal processing techniques with machine learning, our model represents a significant advancement in the medical field, offering a more efficient and accurate method for diagnosing migraines and improving patient care.