This study investigates the diagnostic sensitivity of gamma/beta and alpha/gamma ratios in discriminating dementia patients from healthy subjects. Our analysis uses EEG based Multivariate Fourier Decomposition method to extract the features of brain rhythms (alpha, gamma and beta) and their respective ratios for characterization. These features are analyzed using correlation in the sub-band of the envelope. Based on the results of this study, we propose that features gamma/beta and alpha/gamma ratios can be further analyzed using an extension of this method to fully evaluate their potential in the characterization of dementia. © 2016 IEEE.