The human brain is a complex network of anatomically segregated groups of functionally specialized neurons and functional links between these groups. Thereby, network science can open new insights on the human cognition. However, this requires understanding on how preprocessing of the data and decisions made when constructing the network can affect the results.
Definition of nodes affects properties of functional brain networks. However, there is no clear consensus among neuroscientists on how these nodes should be defined. We aim to present an optimal node definition strategy for networks constructed from functional magnetic resonance imaging (fMRI) data. Such a strategy enables producing reliable and realistic knowledge on the human brain. Since a typical brain network analysis includes combining measurement voxels into larger brain areas, we pay special attention on this coarse-graining process. In order to investigate the effects of node definitions, we introduce consistency as a measure of the internal homogeneity of a brain area. We investigate how the consistency varies between areas and conditions (task versus rest) and as if it can predict the functional role of a brain area.
Spatial smoothing is used as a standard part of the fMRI preprocessing pipeline. However, the effects of spatial smoothing on the analyses of functional brain networks have not been studied in detail. We investigate, which are the cases where spatial smoothing can cause untrivial changes in the structure of the functional brain network and what should one do in order to avoid contamination of analysis results by spatial smoothing.
The functional brain network is constructed by thresholding the correlation matrix of voxel time series. Node definitions, coarse-graining, and spatial smoothing can affect the structure of the network.
Prof. Jari Saramäki
Onerva Korhonen (PhD Student)
Tuomas Alakörkkö (Master Student)