Migraine Analysis with Cross Entropy Based Connectivity Feature: Investigation of Sensory Stimulus Conditions

Migraine is one of the most common neurological disorders. Despite its high prevalence, the lack of research on migraine compared to other neurological disorders is striking. Visual or auditory sensory stimuli are effective in triggering migraine attacks. The relationships between different brain regions depending on the stimuli can be analyzed with brain connectivity properties by creating a connectivity matrix. In this study, the effect of sensory stimuli in migraine patients is analyzed by entropy-based connectivity analysis and machine learning algorithms Support Vector Machine (SVM) and Artificial Neural Network (ANN) algorithms. At the same time, the classification process was carried out with a healthy control group according to the stimulus conditions in order to contribute to the development of systems that can help diagnose migraine. Unlike previous studies, calculations were made between electrode pairs by applying permutation and conditional Cross Entropy (CE) techniques to Electroencephalography (EEG) signals and connectivity feature matrix or map showing the connection between electrodes was created. In the analysis of the stimulus effect in migraine patients, the most successful classification was obtained between resting and auditory stimulus conditions in the ANN algorithm (Accuracy: 83.68%). In the classification of migraine patients with healthy control group, the ANN algorithm and auditory stimulus condition gave the most successful accuracy rate (Accuracy: 85.71%). According to the analyses in this study, it was determined that the auditory stimulus condition may show significant differences in certain channels in certain regions of the brain of migraine patients compared to visual stimulus and resting conditions and healthy participants. The proposed approach (Fused CE+ANN) has shown successful performance in analyzing the stimulus effect and predicting migraine in migraine patients. Cross entropy techniques can help to discover brain connectivity features that may occur in the brain activity information that may occur especially under stimulus effect.