Alpha Frequency Estimation in Patients With Epilepsy

Abstract

We report comparison and assessment of the clinical utility of different automated methods for the estimation of the alpha frequency in electroencephalograph (EEG) and compare them with visual evaluation. A total of 56 consecutive patients, aged 17 to 78 years, who had a routine EEG recording, were included, and they were grouped as patients with epilepsy (Ep) and without epilepsy (nEp). Five different methods were used for alpha frequency estimation: visually guided manual counting and visually guided Fourier transform, and 3 methods were fully automated: time domain estimation of alpha (automatic assessment of alpha waves in time domain [ATD]) and 2 fast Fourier transform (FFT)-based methods, a segmented (automatic assessment of EEG segments by FFT) and one full FFT (automatic assessment of whole EEG by one FFT of the full recording [AWF]). The AWF discriminated significantly between Ep and nEp. Visually guided manual counting showed an almost significant difference independently in the 2 occipital electrodes. The ATD underestimated high frequencies and returned a too low mean frequency. This study shows that AWF is the best suited method for automatic assessment of the alpha frequency.