Cortical folding patterns and predicting cytoarchitecture

Abstract

The human cerebral cortex is made up of a mosaic of structural areas, frequently referred to as Brodmann areas (BAs). Despite the widespread use of cortical folding patterns to perform ad hoc estimations of the locations of the BAs, little is understood regarding 1) how variable the position of a given BA is with respect to the folds, 2) whether the location of some BAs is more variable than others, and 3) whether the variability is related to the level of a BA in a putative cortical hierarchy. We use whole-brain histology of 10 postmortem human brains and surface-based analysis to test how well the folds predict the locations of the BAs. We show that higher order cortical areas exhibit more variability than primary and secondary areas and that the folds are much better predictors of the BAs than had been previously thought. These results further highlight the significance of cortical folding patterns and suggest a common mechanism for the development of the folds and the cytoarchitectonic fields.

Figure 1.

Image of a representative reconstructed histology dataset with the gray/white surface overlaid in yellow. Top: coronal, middle: sagittal, bottom: axial. The green crosses represent the locations of the manually selected control points in the body of the white matter for these slices. The red cross represents the point in common for the 3 views.

Figure 2.

Spatial probability maps of different BAs. Top row: left hemisphere areas 17, 18, 4p, and 2 (from left to right). Second row: areas 4a, 6, 44, and 45. Third and fourth rows are same as first and second for the right hemisphere. Fifth row: color scale used for spatial probability maps.

Figure 3.

Cumulative histograms of surface (white) and nonlinear volume (black) accuracy. From left to right: top row: V1 and V2, second row: 4p and 2, third row: 4a and 6, fourth row: 44 and 45. Bottom: mean across the set of areas (left), black bars are standard errors of the mean, and ratio of surface to volume (right).

Figure 4.

Box plot of the accuracy of the predicted location of the borders of a set of BAs as estimated by the average distance between boundaries of the corresponding areas in millimeter, using a modified Hausdorff distance (y-axis). The higher the accuracy, the lower the distance. Left: 8 BAs in the left hemisphere, right: 8 BAs in the right hemisphere. Blue boxes indicate the lower and upper quartiles; red lines are the medians. The lines extend to 1.5 times the interquartile spacing. Data points outside of the lines are outliers.