Abstract
In accordance with the concept of topographic organization of neuroanatomical structures, there is an increased interest in estimating and delineating continuous changes in the functional connectivity pattern of neighboring voxels across a region using resting-state fMRI data. Fundamental to this topographic connectivity estimation is the assumption that the organizational gradient is stable across the region of interest. If, however, the region of investigation is actually comprised of multiple subdivisions with different gradients, these estimations would be misleading. Here, we present a model testing procedure to arbitrate between overlapping, shifted, or different topographic connectivity gradients across subdivisions of a structure. We found that this model selection procedure performed well in control conditions of checkerboard subdivisions, which are spatially overlapping subsamples of a region, demonstrating support for overlapping gradients. We then applied the procedure to the striatum, a subcortical structure consisting of the caudate nucleus and putamen, in which an extensive literature suggest to have a shared topographic organization of a diagonal gradient. We found, across multiple resting state fMRI data samples of different spatial resolutions in humans, and one macaque resting state fMRI data sample, that the model with different connectivity gradients across the caudate and putamen was preferred across all data samples. More specifically, within the caudate, we replicated the diagonal organization, but also revealed a medial-lateral organization. Clustering of the putamen only revealed the diagonal gradient. These findings demonstrate the importance of testing basic assumptions even in the case of strong literature support. Indeed, preforming the same analysis assuming a unitary gradient obfuscates the medial-lateral organization of the caudate. Further studies are needed to determine the importance of differential connection gradients across the putamen and caudate and the medial-lateral gradient of the caudate in humans.