Effect of Flash Stimulation for Migraine Detection Using Decision Tree Classifiers
Abstract Migraine is one of the most persistent neurological disorder in the world. An effective migraine diagnosis is seen as elusive and deeply problematic, since migraine covers a broader range of different symptoms. Hence, designing a computer-assisted migraine diagnosis system is still an open research area. Migraine subjects (MSs) show apparent discrepancies in pain-related evoked responses as well as decreased adaptation to continuous repeating stimulation. In the present contribution, we suggest an innovative analysis for EEG signal band synchronization, along direct dynamic impacts under painful stimuli in MSs, and compare the results with non-migraine control subjects. The main aim of this contribution is to evaluate the impact of flash stimulation on classification performances and to find the effective length of window in classification of EEG signals. In this paper, decision tree-based classifier system is designed for migraine diagnosis. In the first step, we decompose EEG signals decomposed into different frequency sub-bands by using DWT (discrete wavelet transform). In the next step, we extract different statistical features from DWT sub-bands. In the last step, we feed the features into decision tree classifier. Experimental results show that flash stimulation affects the classification accuracy. Additionally, the window length affects the classification accuracy as well.