Migraine detection through spontaneous EEG analysis

Spontaneous EEG patterns are studied to detect migraine patients both during the attack and in headache-free periods. The EEG signals are analyzed through the wavelets and both scale-dependent and scale-independent features are computed to characterize the patterns. The classification is carried out by a supervised neural network. The efficiency of the method is evaluated through the Receiver Operating Characteristic (ROC) analysis and the Wilcoxon-Mann-Whitney (WMW) test. Although a high discrimination is observed with one single neural output, a complete separation among MwA patients and healthy subjects is obtained when a scatter plot is drawn in the plane of two suitable neural outputs. © 2007 IEEE.