Book Description
Functional magnetic resonance imaging (fMRI) based on blood-oxygen level dependent (BOLD) contrast is a powerful technique for non-invasive measurement of brain activity. Recent fMRI studies have revealed that the spontaneous BOLD fluctuations of the human brain organize into distributed, temporally-coherent networks ("resting-state networks"; RSNs). Examination of RSNs has yielded valuable insight into neural organization and development, and demonstrates potential as a biomarker for conditions such as Alzheimer's disease and depression. However, the accuracy by which the spatio-temporal properties of RSNs can be delineated using fMRI is compromised by the presence of physiological (cardiac and respiratory) noise and vascular hemodynamic variability. Further, our present understanding of how RSNs may interact and support cognitive function has been limited by the fact that the vast majority of studies to-date analyze RSNs in a manner that assumes temporal stationarity. Here, we describe efforts to correct for non-neural physiological influences on the BOLD signal, as well as investigations into the dynamic character of resting-state network connectivity. It is found that low-frequency variations in cardiac and respiratory processes account for significant noise across widespread gray matter regions, and that a constrained deconvolution approach may prove effective for modeling and reducing their effects. Application of the proposed noise-reduction procedure is observed to yield negative correlations between the spontaneous fluctuations of two major RSNs. The relationship between respiratory volume changes and the BOLD signal is further examined by simultaneously monitoring and comparing chest expansion data, end-tidal gas concentrations, and spontaneous BOLD fluctuations. The use of a breath-holding task is proposed for quantifying regional differences in BOLD signal timing that arise from local vasomotor response delays; such non-neural timing delays are found to impact inferences of resting-state connectivity and causality. Finally, a preliminary analysis of non-stationary connectivity between RSNs is performed using wavelet and sliding-window approaches, and it is observed that interactions between networks may reconfigure on time-scales of seconds to minutes.
