A multi-modal parcellation of human cerebral cortex

Abstract

Understanding the amazingly complex human cerebral cortex requires a map (or parcellation) of its major subdivisions, known as cortical areas. Making an accurate areal map has been a century-old objective in neuroscience. Using multi-modal magnetic resonance images from the Human Connectome Project (HCP) and an objective semi-automated neuroanatomical approach, we delineated 180 areas per hemisphere bounded by sharp changes in cortical architecture, function, connectivity, and/or topography in a precisely aligned group average of 210 healthy young adults. We characterized 97 new areas and 83 areas previously reported using post-mortem microscopy or other specialized study-specific approaches. To enable automated delineation and identification of these areas in new HCP subjects and in future studies, we trained a machine-learning classifier to recognize the multi-modal 'fingerprint' of each cortical area. This classifier detected the presence of 96.6% of the cortical areas in new subjects, replicated the group parcellation, and could correctly locate areas in individuals with atypical parcellations. The freely available parcellation and classifier will enable substantially improved neuroanatomical precision for studies of the structural and functional organization of human cerebral cortex and its variation across individuals and in development, aging, and disease.

Figure 1. Consistency of fine spatial details in independent group averages.

Relative myelin content maps (left hemisphere) and task fMRI contrast beta maps from the LANGUAGE story contrast (right hemisphere) on inflated (columns 1 and 3) and flattened surfaces (columns 2 and 4). Rows 1 and 2 are the group averages of the 210P and 210V data sets, respectively. White and black arrows indicate consistent variations in myelin content within primary somatosensory cortex that are correlated with somatotopy (see Supplementary Neuroanatomical Results 6 and Supplementary Neuroanatomical Results Fig. 8). The white oval indicates a small, sharp, and reproducible feature in the right hemisphere of the LANGUAGE story contrast. Relative myelin content will hereafter be referred to as myelin (see legend of Supplementary Fig. 1 in Supplementary Results and Discussion 1.2). Data at (http://balsa.wustl.edu/WDpX).

Figure 2. Parcellation of exemplar area 55b using multi-modal information.

The border of 55b is indicated by a white or black outline. a, Myelin map. b, Group average beta map from the LANGUAGE Story versus Baseline task contrast. c, d, Functional connectivity correlation maps from a seed in area PSL (white sphere, arrow) (c) and a seed in area LIPv (white sphere, arrow) (d). e, Gradient magnitude of the myelin map shown in a. f, Gradient magnitude of the LANGUAGE Story versus Baseline task contrast shown in b. g, Mean gradient magnitude of the functional connectivity dense connectome (see section on modalities for parcellation in the Methods). h, A dorsal schematic view of the prefrontal cortex as parcellated in ref. 22, in which shading indicates the amount of myelin found using histological stains of cortical grey matter. Data at (http://balsa.wustl.edu/Qv4P).

Figure 3. The HCP’s multi-modal parcellation, version 1.0 (HCP_MMP1.0).

The 180 areas delineated and identified in both left and right hemispheres are displayed on inflated and flattened cortical surfaces. Black outlines indicate areal borders. Colours indicate the extent to which the areas are associated in the resting state with auditory (red), somatosensory (green), visual (blue), task positive (towards white), or task negative (towards black) groups of areas (see Supplementary Methods 5.4). The legend on the bottom right illustrates the 3D colour space used in the figure. Data at http://balsa.wustl.edu/WN56).

Figure 4. Example parcellated analyses using the HCP’s multi-modal cortical parcellation.

a, Dense myelin maps on lateral (top) and medial (bottom) views of inflated left hemisphere. b, c, Example dense (b) and parcellated (c) task fMRI analysis (LANGUAGE story versus baseline) expressed as Z statistic values. d, The entire HCP task fMRI battery’s Z statistics for 86 contrasts (47 unique, see section on modalities for parcellation in the Methods) analysed in parcellated form and displayed as a matrix (rows are parcels, columns are contrasts, white outline indicates the map in c). e, A major improvement in Z statistics from fitting task designs on parcellated time series instead of fitting them on dense time series and then parcellating afterwards (blue points are 360 parcels × 86 task contrasts; note the upward tilting deviation from the red line). f, Parcellated myelin maps. g, A parcellated folding-corrected cortical thickness map (in mm). h, i, Parcellated functional connectivity maps on the brain (seeded from area PGi, black dot). These parcellated connectomes are computed using either full or partial correlation (see Supplementary Methods 7.1). In both cases, the task negative (default mode) network is apparent. j, A parcellated connectome matrix view with the full correlation connectome below and the partial correlation connectome above the diagonal (white line shows the displayed partial correlation brain map). Data at (http://balsa.wustl.edu/RG0x).

Figure 5. Areal detection rates, probabilistic areas, and parcellation reproducibility.

Rows 1 (210P) and 2 (210V) show the individual subject areal detection rates (see Methods section on cortical areal classifier) as parcellated maps. Most areas are yellow (100%), and the minimum detection rate across both rows was 73%. Rows 3 and 4 illustrate probabilistic maps of areas V1, M1, RSC, MT, LIPv, TE1a, 46, and 10r for the 210P (row 3) and 210V (row 4) groups. Row 5 shows the original parcellation derived from the semi-automated neuroanatomical approach. Row 6 shows the group MPM maps from 210P (blue), 210V (red), and their overlap (purple). Data at (http://balsa.wustl.edu/WL8m).