Precise Topology of Adjacent Domain-General and Sensory-Biased Regions in the Human Brain

Abstract

Recent functional MRI studies identified sensory-biased regions across much of the association cortices and cerebellum. However, their anatomical relationship to multiple-demand (MD) regions, characterized as domain-general due to their coactivation during multiple cognitive demands, remains unclear. For a better anatomical delineation, we used multimodal MRI techniques of the Human Connectome Project to scan subjects performing visual and auditory versions of a working memory (WM) task. The contrast between hard and easy WM showed strong domain generality, with essentially identical patterns of cortical, subcortical, and cerebellar MD activity for visual and auditory materials. In contrast, modality preferences were shown by contrasting easy WM with baseline; most MD regions showed visual preference while immediately adjacent to cortical MD regions, there were interleaved regions of both visual and auditory preference. The results may exemplify a general motif whereby domain-specific regions feed information into and out of an adjacent, integrative MD core.

Keywords: auditory working memory; domain-general; multiple-demand; sensory-biased; visual working memory.

Figure 1

(a) The extended MD system. Core MD regions are colored in dark green and white labels. Penumbra MD regions are in light green with black labels. Note here we separated core region SCEF/8BM (as identified in Assem et al. 2020) into SCEF as penumbra and 8BM as core for simplicity in analysis. (b) and (c) are adapted from Tobyne et al. (2017). (b) Sensory-biased regions [originally identified in (Michalka et al. 2015)] after their transformation to the HCP fs_LR surface. Red: visually biased. Blue: auditory biased. Overlapping HCP MMP1.0 regions are labeled and their original Michalka et al. (2015) labels are in brackets. Green contours correspond to extended MD borders in (a). (c) Sensory-biased lateral frontal regions based on their intrinsic rfMRI connectivity with posterior cortical areas. Black contours surrounding regions with warmer colors (red) are significantly more connected with visual parietal areas than auditory temporal regions. Black contours surrounding regions with colder colors (blue) are significantly more connected with auditory temporal regions than visual parietal regions. Data available at https://balsa.wustl.edu/6VPVv.

Figure 2

N-back task paradigm. Illustration of a stimulus sequence from the hard (3-back) version. Subjects performed a visual and an auditory version of the n-back task in separate sessions. Each task had easy (1-back, not shown) and hard (3-back) blocks. Each block (30 s) started with a cue (4 s) followed by 12 trials (2 s each) and ended with a fixation screen (2 s). Subjects pressed right for target stimuli, and left for all nontarget stimuli. For the visual n-back, stimuli consisted of pictures of houses (illustrated) or faces (not shown). Auditory n-back stimuli consisted of animate (illustrated) or inanimate (not shown) sounds. See Materials and Methods for further details.

Figure 3

Hard > easy n-back activation maps. (a) MD group (n = 449) average activation of 3 cognitively demanding contrasts from Assem et al. (2020). (b) Visual and (c) auditory group (n = 37) average n-back hard>easy activations. (d) Two exemplar single subject activation maps for the hard > easy visual (left) and auditory (right) n-back task. All activation values are percent signal change. Data available at https://balsa.wustl.edu/1BkBG and all 37 subject activation files can be downloaded from https://balsa.wustl.edu/5XxX1.

Figure 4

Hard > easy contrast activations (% signal change) in right (top) and left (bottom) hemispheres, separately for auditory (black) and visual (pink) tasks. Error bars represent standard error of the mean (SEM). Extended MD regions labels are colored in green (core MD in dark green and bold, penumbra MD in light green). The 4 MMP1.0 regions partially overlapping with Michalka et al. (2015) sensory-biased regions are colored in red (visual-biased) and blue (auditory-biased). Asterisks denote P < 0.05 Bonferroni corrected for n = 64 regions.

Figure 5

Visual versus auditory task preferences for the hard > easy contrast. (a) Bar heights represent average activation differences (%signal change, visual hard > easy minus auditory hard > easy) for each region across subjects. Error bars represent SEM. Light colored bars represent regions of the right hemisphere. Areal names are colored in dark bold green (core MD), light green (penumbra MD), red (Michalka-visual), blue (Michalka-auditory) (b) Cortical activations (% signal change) for the same contrast. Black contours surround significant vertices (FDR corrected P < 0.05), gray contours correspond to the HCP MMP 1.0 areal borders and green contours correspond to extended MD areal borders. Data available at https://balsa.wustl.edu/npmpr.

Figure 6

Visual versus auditory task preferences for the easy > fix contrast. (a) Cortical activations (%signal change) for the visual easy > fix minus auditory easy>fix contrast. (b) Gradient map (i.e., 1st spatial derivative) for this contrast. Warmer colors highlight cortical regions with sudden shifts in task modality preferences. Note how the strongest gradients lie just outside core MD regions (white borders). Data available at https://balsa.wustl.edu/g7V7B including all 37 subject activation files for the contrast in (a).

Figure 7

(a) Bar heights represent average activations (% signal change) for each region across subjects for the visual versus auditory easy > fix contrasts. Circles above bars denote P < 0.05 Bonferroni corrected for n = 64 regions. All remaining details are the same as in Figure 5a. (b) left: cyan borders surround the significant vertices (top 20%) within each of the HCP MMP1.0 regions overlapping with Michalka et al regions (data are available at https://balsa.wustl.edu/NwV0p). Right: average activations (% signal change) for each region across subjects for the same contrast. Circles above bars denote P < 0.05 Bonferroni corrected for n = 10 regions.

Figure 8

Subcortical and cerebellar MD visual versus auditory task preferences. (a) Subcortical and cerebellar MD masks identified in Assem et al. 2020. The cerebellum is displayed as a flat surface with black contours representing anatomical borders. Data are available at https://balsa.wustl.edu/MxzxD  (b) Hard > easy activations (% signal change) for the visual and auditory tasks. Left (MD caudate), middle (MD thalamus), right (MD cerebellum). Error bars are SEM. Asterisks denote significant hard>easy activations (p < 0.05; Bonferroni corrected n = 6 regions). n.s. denotes nonsignificant differences between visual hard > easy and auditory hard > easy (P < 0.05; Bonferroni corrected n = 6 regions). (c) Visual versus auditory easy task activations (% signal change) of subcortical and cerebellar MD regions (i.e., easy > fix visual—easy > fix auditory). (d) Left and middle: Subcortical voxels showing significant (FDR P < 0.05) visual versus auditory task activations during easy > fix contrast. Right: cerebellar voxels activations for the same contrast. Significant cerebellar regions are surrounded by white contours. MD regions are delineated by green contours. Data are available at https://balsa.wustl.edu/B4K41.