Triple-Network (TN) and Language Network (LN) Multi-Scale Parcellations

Description: Parcellations of Triple and Language Networks derived from bootstrapped individual + group K-Means clustering. Triple-Network (TN) A connectome can be defined anatomically (i.e. through anatomical atlases) or functionally (i.e. through clustered parcellation), and at multiple resolutions ranging from nodes as individual voxels to nodes as populations of contiguous voxels. To explore the impact of multiple node resolutions while ensuring aligned vertices across subjects, we first used Shaeffer/Yeo-7 multi-resolution anatomical atlas, which provides parcellations for the three "triple" resting-state networks of attention (DMN, ECN, SN), a subset of which collectively constitute what we will refer to as the Triple-Network (TN). The resulting parcels spanned the following neuroanatomical labels: Frontal (Ventral Medial, Lateral Inferior, Superior-Medial), Cingulate (Anterior-Superior, Subgenual Anterior, Middle, and Posterior), Temporal (Inferior, Anterior Superior), Parietal (Middle Inferior, Lateral Superior), Right Precuneus, Insula (Anterior), Precentral Gyri, Intraparietal Sulcus, and Bilateral Supramarginal Gyrus. To contrastingly define the intersection of the TN, we further constructed a study-specific parcellation using bootstrapped K-Means spectral clustering with spatial constraints to only those voxels defined at the "zones of intersection" of the TN as defined by a spatial mask derived from the Shaeffer/Yeo parcellations. To facilitate meaningful comparison, we produced a similar number of clusters for the intersection case as for the union (though the former covered fewer cumulative voxels), and included primarily the following sub-set of structures: Anterior, Middle, and Posterior Cingulate, Right Precuneus, Anterior Insula, Bilateral Precentral Gyri, Intraparietal Sulcus, and Bilateral Supramarginal Gyri. Language-Network (LN) To generate a parcellation that would also carry utility for multimodal analysis, however, we again used bootstrapped K-Means clustering of each subject's resting-state data to define parcels. This time, however, we constrained clustering to only those voxels contained within anatomically defined language ROI's, the majority of which were obtained from the Harvard-Oxford subcortical atlas, but all of which were manually resampled to the CN200 template. These included the Broca's Area, Wernicke's Area, Planum Temporale, Posterior Inferior Temporal Gyrus, Posterior Superior Temporal Gyrus, Posterior Supramarginal Gyrus, Middle Temporal Gyrus, Insular Cortex, and the Precentral Gyrus. As identified by "connectional" cortical atlas, those neuroanatomical labels in common across all network definitions included: Cingulate Gyrus (CG),Precuneus (Pcun), Insula (INS), Inferior Frontal Gyri (IFG), Inferior Parietal Lobule (IPL), Middle Frontal Gyrus (MFG), Middle Temporal Gyrus (MTG), Orbital Gyrus (OrG), and Superior Frontal Gyrus (SFG).