Principles of temporal association cortex organisation as revealed by connectivity gradients

Abstract

To establish the link between structure and function of any large area of the neocortex, it is helpful to identify its principles of organisation. One way to establish such principles is to investigate how differences in whole-brain connectivity are structured across the area. Here, we use Laplacian eigenmaps on diffusion MRI tractography data to investigate the organisational principles of the human temporal association cortex. We identify three overlapping gradients of connectivity that are, for the most part, consistent across hemispheres. The first gradient reveals an inferior-superior organisation of predominantly longitudinal tracts and separates visual and auditory unimodal and multimodal cortices. The second gradient radiates outward from the posterior middle temporal cortex with the arcuate fascicle as a distinguishing feature; the third gradient is concentrated in the anterior temporal lobe and emanates towards its posterior end. We describe the functional relevance of each of these gradients through the meta-analysis of data from the neuroimaging literature. Together, these results unravel the overlapping dimensions of structural organization of the human temporal cortex and provide a framework underlying its functional multiplicity.

Keywords: Connectivity gradients; Laplacian eigenmapping; Temporal lobe; Tractography.

Fig. 1

Group-averaged connectivity modes projected on a mid-thickness surface: a Group-averaged dominant mode of connectivity (g1), b group-averaged second dominant mode of connectivity (g2), c group-averaged third dominant mode of connectivity (g3) Left left hemisphere, Right right hemisphere. Similar colours represent similar gradient values

Fig. 2

Projection images of a representative subject—103818: a Projection image of the dominant mode of connectivity (g1), b projection image of the second dominant mode of connectivity (g2), c projection image of the third dominant mode of connectivity (g3), Left left hemisphere, Right right hemisphere. The scale is normalised between 1 and 10 and similar colours represent similar gradient values along the represented connectivity mode

Fig. 3

Projection image. Squares (left hemisphere) and diamonds (right hemisphere) represent average values of the temporal lobe’s relevant tract projections along the first two gradients. Dashes represent the bootstrapped 95% confidence interval of the mean. X-axis—value along the projection image of the first mode of connectivity (g1; from interior to superior positions); Y-axis—value along the projection image of the second mode of connectivity (g2; radiates from the AF).

Fig. 4

Projection image. Squares (left hemisphere) and diamonds (right hemisphere) represent average values of the temporal lobe’s relevant tract projections along the first and third gradients. Dashes represent the bootstrapped 95% confidence interval of the mean. X-axis—value along the projection image of the first mode of connectivity (g1; from inferior to superior positions); Y-axis—value along the projection image of the third mode of connectivity (g3; from anterior to posterior positions).

Fig. 5

Functional decoding tables of the dominant connectivity gradient of the left (a) and right (b) temporal lobe. On top of each functional decoding table, the gradient deciles used can be seen separated on a mid-thickness surface. On the table, the X-axis represents the decile being decoded and the Y-axis depicts the terms that Neurovault identified as most spatially correlated with the connectivity gradient. Values represent Pearson’s spatial correlation of the decile with the whole brain activation pattern associated with the term

Fig. 6

Functional decoding of the second dominant connectivity gradient of the left (a) and right (b) temporal lobes. On top of each functional decoding table, the gradient deciles used can be seen separated on a mid-thickness surface. On the table, the X-axis represents the decile being decoded and the Y-axis depicts the terms that Neurovault identified as most spatially correlated with the connectivity gradient. Values represent Pearson’s spatial correlation of the decile with the whole brain activation pattern associated with the term

Fig. 7

Functional decoding of the second dominant connectivity gradient of the left (a) and right (b) temporal lobe. X-axis represents the decile being decoded and the Y-axis depicts the terms that Neurovault identified as most spatially correlated with the connectivity gradient. Values represent Pearson’s spatial correlation of the decile with the whole brain activation pattern associated with the term