fMRI Functional Cluster Analysis Using the Stockwell Transform

Classical functional connectivity analysis of resting state fMRI data computed over the length of an entire scan implies stationarity for the signals and neglect changes in connectivity that occur on a much shorted scale. Recently, interest has been growing in dynamic analysis methods that can detect changes in connectivity on orders comparable with the cognitive process. Previous work showed that these changes in functional connectivity can be observed using classical sliding window techniques in human and animal subjects, although the hypothesis of stationarity on the data leads to suboptimal parcellations of the brain and to results that are dependent on the length of the chosen window. Recent techniques based on the wavelet and wavelet packet decomposition and clustering of resting state fMRI data overcome these obstacles with data-driven functional clusters based on temporal and spectral properties of the signals. The use of precise time-frequency techniques is important in the characterization of the dynamic properties of these clusters. The Stockwell transform offers good resolution in the time-frequency retaining at the same time the absolute phase information of the signal. In this work, we propose a study of the time frequency characteristics of wavelet-based functional clusters based on the Stockwell transform.