Datensatz

Zusammenfassung

Introduction:
For understanding proper functioning of the neural circuits, one has to consider dynamical changes in brain activation unfolding over time: distinct networks can be topographically similar, but differ in terms of temporal dynamics. Time is a key factor in the brain network organization. Brain topography alone does not allow us to uncover neural communication in the brain. In the present study, we analyzed temporal dynamics of the BOLD response to point-light body motion (BM). Recent fMRI also focuses on interactions between brain regions making up the social brain (Sokolov et al., 2018). The present analysis was motivated by a desire to characterize the functional role of the brain regions playing in unison at different time points and, thus, making up diverse large-scale circuits.
Methods:
By using whole-brain coverage, we conducted fMRI recording at field strength of 9.4 tesla during processing of point-light BM. Adults were administered a 2-AFC task: they indicated whether an upright walker or control configurations were presented. For uncovering ensembles of regions playing in unison over the whole brain, we used the following three strategies: 1. Temporal contrasts analysis (bin 5 s): Voxel based whole brain bin-pattern search. In our earlier work (Pavlova et al., 2017), the temporal dynamics of the BOLD response was analyzed within several foci of activation during processing of upright and inverted BM. Here the whole brain search for these specific bin-patterns was performed. 2. Independent component analysis (ICA) based on temporal pattern (bin size 1.3s = TR). ICA (Calhoun et al., 2001) was performed on the contrast images of the second model. For decreasing the number of components, in a first data reduction step 30 principle components (PC) were calculated for each subject after removing the mean over all voxel per time point. In a second PCA, for avoiding too many clusters with noise time course, they were collapsed to 6 group components. In the next step, 6 independent group components (IC) were extracted using the Infomax algorithm (Bell and Sejnowski, 1995). 3. Kmeans clustering based on temporal pattern analysis (bin size 5s). The 2 (conditions) x 4 (5s bins) contrast images of the 5s time bins from the first model were used to find cluster of typical time courses by the kmeans clustering method (Lloyd, 1982). The number of clusters was set to K=6 cluster centers and the algorithm was repeated 10 times with different random chosen centroid seeds using the k-means++ algorithm (Arthur and Vassilvitskii, 2007). The solution with the lowest within-cluster sums of point-to-centroid distances was selected for the final outcome.
Results:
All methods resulted in revealing large-scale networks with similar topographical distribution. For example, they point to the network comprising mainly posterior brain areas known to be involved in cognitive processing of the sensory input. In turn, the temporal contrasts analysis and the ICA also reveal the network involving mainly bilateral postcentral and precentral areas, portions of the supplementary motor areas, superior parietal cortex, rolandic operculum, supramarginal areas in both hemispheres and the right insula. This network involves the areas known to be engaged in decision making and readiness to motor response. However, some methods revealed networks that were undetected by other methods. For example, the network that involves mainly anterior brain areas exhibit a remarkable initial decrease in the brain activity during earlier stages of stimulus presentation. These brain areas most likely forming up the default mode network that is usually active without any explicit task. The temporal dynamics of this network is different from or even opposite to other task-engaged circuits.
Conclusions:
The benefits and disadvantages of these methods for unveiling of temporal dynamics of different large-scale networks are discussed. The outcome opens novel perspectives for the social brain assessment.