Large-Scale Meta-Analysis of Human Medial Frontal Cortex Reveals Tripartite Functional Organization

Abstract

The functional organization of human medial frontal cortex (MFC) is a subject of intense study. Using fMRI, the MFC has been associated with diverse psychological processes, including motor function, cognitive control, affect, and social cognition. However, there have been few large-scale efforts to comprehensively map specific psychological functions to subregions of medial frontal anatomy. Here we applied a meta-analytic data-driven approach to nearly 10,000 fMRI studies to identify putatively separable regions of MFC and determine which psychological states preferentially recruit their activation. We identified regions at several spatial scales on the basis of meta-analytic coactivation, revealing three broad functional zones along a rostrocaudal axis composed of 2-4 smaller subregions each. Multivariate classification analyses aimed at identifying the psychological functions most strongly predictive of activity in each region revealed a tripartite division within MFC, with each zone displaying a relatively distinct functional signature. The posterior zone was associated preferentially with motor function, the middle zone with cognitive control, pain, and affect, and the anterior with reward, social processing, and episodic memory. Within each zone, the more fine-grained subregions showed distinct, but subtler, variations in psychological function. These results provide hypotheses about the functional organization of medial prefrontal cortex that can be tested explicitly in future studies.

Significance statement: Activation of medial frontal cortex in fMRI studies is associated with a wide range of psychological states ranging from cognitive control to pain. However, this high rate of activation makes it challenging to determine how these various processes are topologically organized across medial frontal anatomy. We conducted a meta-analysis across nearly 10,000 studies to comprehensively map psychological states to discrete subregions in medial frontal cortex using relatively unbiased data-driven methods. This approach revealed three distinct zones that differed substantially in function, each of which were further subdivided into 2-4 smaller subregions that showed additional functional variation. Each individual region was recruited by multiple psychological states, suggesting subregions of medial frontal cortex are functionally heterogeneous.

Keywords: cognitive control; medial frontal cortex; meta-analysis; pain.

Figure 1.

Methods overview. A, Whole-brain coactivation of MFC voxels was calculated, and k-means clustering was applied, resulting in spatially distinct clusters. B, For each cluster, thresholded whole-brain coactivation maps were generated. C, We generated functional preference profiles for each cluster by determining which psychological topics best predicted their activation.

Figure 2.

Coactivation-based clustering of MFC results. A, Mid-sagittal view at three levels at granularity: three broad zones, nine and 12 subregions. Clusters in nine subregion solution are given both descriptive and alphanumeric names for reference. B, Axial view of nine subregions. C, Silhouette scores of real (green) and permuted (blue) clustering solutions. Clustering was performed on permuted data 1000 times for each k to compute a null distribution (p values for all clusters < 0.001). Silhouette scores reached local maxima at 3 regions and plateaued after 9.

Figure 3.

Meta-analytic coactivation contrasts for (A) three zones and (B) nine subregions. Colored voxels indicate significantly greater coactivation with the seed region of the same color (at right) than control regions in the same row. The three zones showed distinct coactivation patterns, whereas subregions within each zone showed fine-grained coactivation differences. Images are presented using neurological convention and were whole-brain corrected using a false discovery rate of q = 0.01. Major subcortical structures are labeled as follows: Thal, Thalamus; Hipp, hippocampus; Amyg, amygdale; VS, ventral striatum; DS, dorsal striatum.

Figure 4.

Functional preference profiles of MFC clusters. Each cluster was profiled to determine which psychological concepts best predicted its activation. Top, Each of the three functional zones we identified showed distinct functional profiles with broad shifts across cognitive domains. Bottom, Within each zone, subregions showed fine-grained shifts in function. Strength of association is measured in LOR, and permutation-based significance (p < 0.001) is indicated next to each psychological concept by color-coded dots corresponding to each region.