Spatial navigation is a complex process that involves a distributed network in the brain. Our work presents an integrative approach for modeling a functional network for spatial navigation using FMRI data. The optimal network model provides new insights for understanding how brain regions within the network interact.
The network model could also better capture variability in behaviors in both health and disease. Our method can also be applied to studies or other complex functions. For instance, a similar procedure could be used for modeling brain network for language and memory.
To begin, check the data quality and exclude participants with missing retest data and excessive head motion. Then open the graph theoretical network analysis or GRETNA toolbox in MATLAB. Click the tab FC Matrix Construction"and select the path of the functional dataset to load the nifty documents.
Execute the steps as displayed in the pipeline option. Next, to perform network node definition, download the latest Neurosynth database by typing the following command. Then generate a new dataset instance from database.
txt and add features to this data by typing this command. Run a meta analysis of the term of interest such as navigation by typing the command. Next, define clusters of interest by incorporating the meta analytic map and the whole brain parcelization atlas such as AICHA or AAL by typing this command from FSL.
Then type the following scripts in Python to check the size of each region in the map. After that, integrate all brain regions into a template by FSL Maths in FSL. For network connectivity estimation, click on FC Matrix Construction in the GRETNA software and click Static Correlation.
Upload the node obtained from network node definition as an atlas to calculate the static correlation of rs-fMRI signals of each pair of regions and transfer them into Fisher's Z scores. To get a positive and weighted network with the steps displayed in the pipeline option, click on Network Analysis. Then add the network matrices into the brain connectivity matrix window and choose an output directory for preparation.
Add a small world, global efficiency, clustering coefficient, shortest path length, degree centrality, and local efficiency to the GRETNA network metric analysis pipeline. Select positive in the sign of matrix. Select network sparsity in the thresholding method and enter a set of threshold sequences.
Choose the network type as weighted. Set the random network number as 1, 000 and click on Run. To determine the optimal number of modules in the network, first calculate the averaging navigation network.
Then divide the resulting average network into 2, 3, 4, and 5 modules using the function, spectral cluster, in MATLAB. Then align the module divisions using the script, procrustes_alignment. m, and calculate the portion of nodes divided into the same module in rest one and rest two.
Select the number of modules with the highest repeatability. To perform the network analysis, examine the similarity of these network metrics between two networks with different types of strategies for no definition like NaviNet AICHA and NaviNet AAL. Check the test retest reliability of these network metrics using the function ICC in MATLAB.
In this study, 27 brain regions associated with spatial navigation were identified using the AICHA atlas. These regions consisted of the medial temporal and parietal regions that have been reported in navigation neuroimaging studies. As a comparison, 20 regions from the AAL atlas were included.
It showed a large overlap and similar community distribution between the two sets including similar ventral and dorsal modules in both networks. Further, five of the six metrics, except for the clustering coefficient, showed significant correlations between the two networks. The similarity values increased with the sparsity threshold for almost all metrics, suggesting that the network level analysis could reflect stable individual differences, independent of node definition choices, and that the sparsity threshold of 0.30 to 0.40 would result in better generalizability in navigation network analysis.
Moreover, evaluation of the rest retest reliability of the topological measures of the navigation networks indicated that the majority of the network's metrics showed fair to good reliability in AICHA, while the AAL network showed relatively higher reliability. In addition, including global signal regression in FMRI data pre-processing could result in high reliability. These results suggested that the clustering coefficient and small world are the most reliable among these metrics.
With this approach, researchers can investigate the developmental trajectory of functionally specific networks. Network properties also provide new biomarkers for guiding early identification of brain disorders such as Alzheimer's disease.
