Identifying Basal Ganglia divisions in individuals using resting-state functional connectivity MRI

Abstract

Studies in non-human primates and humans reveal that discrete regions (henceforth, "divisions") in the basal ganglia are intricately interconnected with regions in the cerebral cortex. However, divisions within basal ganglia nuclei (e.g., within the caudate) are difficult to identify using structural MRI. Resting-state functional connectivity MRI (rs-fcMRI) can be used to identify putative cerebral cortical functional areas in humans (Cohen et al., 2008). Here, we determine whether rs-fcMRI can be used to identify divisions in individual human adult basal ganglia. Putative basal ganglia divisions were generated by assigning basal ganglia voxels to groups based on the similarity of whole-brain functional connectivity correlation maps using modularity optimization, a network analysis tool. We assessed the validity of this approach by examining the spatial contiguity and location of putative divisions and whether divisions' correlation maps were consistent with previously reported patterns of anatomical and functional connectivity. Spatially constrained divisions consistent with the dorsal caudate, ventral striatum, and dorsal caudal putamen could be identified in each subject. Further, correlation maps associated with putative divisions were consistent with their presumed connectivity. These findings suggest that, as in the cerebral cortex, subcortical divisions can be identified in individuals using rs-fcMRI. Developing and validating these methods should improve the study of brain structure and function, both typical and atypical, by allowing for more precise comparison across individuals.

Keywords: connectome; cortico-basal ganglia loops; functional connectivity; graph theory; striatum.

Figure 1

(A) Flowchart of analysis stream. (B) Time courses extracted from two basal ganglia voxels ([−11 5 12] and [11 5 12]) are highly correlated (r = 0.70). Time courses such as these were used to generate whole-brain correlation maps for each basal ganglia voxel.

Figure 2

Rows 1–3. From Cohort One, three subjects’ basal ganglia voxels colored with respect to modularity optimization groupings (shown on each subject's MP-RAGE; coloring for each hemisphere and each subject is arbitrary). Arrows indicate modules labeled as dorsal caudate (red arrows, z = 16), dorsal caudal putamen (blue arrows, z = 10), and ventral striatum (purple arrows, z = −8). Row 4. Conjunction of modules ascribed the same label across Cohort One subjects. Color bar depicts number of subjects with a module assignment at each voxel.

Figure 3

Z-transformed rs-fcMRI maps from modularity assignments are statistically reliable within each cohort for the left hemisphere divisions (first and second rows, z > 3. 00, k = 21, corresponding to p < 0.05, Monte Carlo corrected) and yield common regions of correlation across cohorts (conjunction analysis, third row). Positive correlations are depicted in warm colors (first two rows) and their overlap is depicted in red in the conjunction analysis (third row). Negative correlations are depicted in cool colors (first two rows) and their overlap is depicted in green in the conjunction analysis (third row).

Figure 4

Z-transformed rs-fcMRI maps from three representative subject's modularity optimization assignments for the left hemisphere (putative dorsal caudate, left column, putative dorsal caudal putamen, middle column, and putative ventral striatum, right column, z > 2. 00) are similar. Black circles depict regions identified from the random-effects analysis (superior frontal gyrus: lateral rendering, first column; anterior cingulate cortex, medial rendering, first column; ventral premotor cortex: lateral rendering, second column; pre-supplementary motor cortex: medial rendering second column; orbitofrontal cortex: ventral rendering, third column). Row 4. Conjunction image of all subjects rs-fcMRI maps (z > 2.00).