Estimation of Intrinsic Connectivity Networks by Multivariate Empirical Mode Decomposition

Functional brain mapping is based on electrical and haemodynamic changes occured in the brain. Blood oxeygenated level dependency (BOLD) signal can be non-invasively collected through the use of functional Magnetic Resonance Imaging (fMRI). Brain functional activity can also be observed in the absence of a given task. These activation patterns are named as brain resting state networks. The aim of this study is, to perform functional brain mapping using the coherence metrics between the decomposed BOLD time series insted of using the raw BOLD time series. Multivariate Emprical mode decomposition procedure is applied for the BOLD series decomposition. Limited number of anatomical locations are selected for node positions using anatomical templates. Further each subseries are used to compute the correlations in frequency domain as coherence values between the node points. By this way, spectral properties of subseries are investigated without imposing any a priori information. FMRI data were collected from 19 volunteers and the preprocessing steps are applied prior to analysis of spectral properties. Four subcomponents whose spectral peaks are determined at 0.007 Hz, 0.014 Hz, 0.03 Hz and 0.064 Hz were determined. In the first component, superior temporal gyrus and occipital lobe connections were exhibited which contribute to the functionality of the auditory and visual networks. Posterior and anterior cingulate areas that are the major parts of the default mode network were found to be present in the second component. In the third component, nodes of the attention network were observed with a center frequency of 0.03 Hz to 0.06 Hz. Additionally, connections of superior temporal gyrus were observed in the fourth component