Identifying the Spectral-Based Neurophysiological Biomarkers to Detect Panic Disorder from Alpha Band Using Machine Learning Algorithms

Panic Disorder (PD) is a debilitating condition marked by sudden, intense fear episodes with physical symptoms. Swift and accurate PD detection is crucial for effective intervention. This study aimed to propose an optimal combination of spectral features of the Alpha band to detect PD. For this purpose, 21 PD-diagnosed individuals and 26 healthy controls attended a 5-minute eyes-closed resting state Electroencephalography (EEG) recording session. Welch method was applied to calculate the power spectral density of EEG signals and then the sum, average, maximum, relative power of alpha band, and individual alpha frequency (IAF) were extracted. Relief and nearest component analysis (NCA) methods were performed to select highly relevant features. The maximum average accuracy was reached when commonly selected features between two selection methods were used as inputs of classifiers. Adaboost classifier reached the highest average accuracy with $89.03\%\pm 6.73\%$ rate.