Caret is a free program for surface-based analysis and visualization of brain data, developed by David van Essen and colleagues. It allows for the segmentation, surface reconstruction, flattening, and visualization of cortical surface, including functional activations. In my research, I found it useful, not only to visualize my data on a surface-based representation, but also to do statistics of functional activations on those surfaces. With this motivation, I developed a Matlab-toolbox, which allows reading, writing and manipulating caret-files. These functions provide functionality for performing t-tests and simple linear models (ANOVA's) on a surface. A table of local maxima, with corrected voxel- and cluster-wise p-values (using Gaussian Field theory, following exactly the model established in SPM2) can then be easily generated.

 

STATISTICS: one-sample t-test: mean<0

=======================================================

c      NumN   Area   max(Z) p(unc) p(cor) p(cl.) X(mm)  Y(mm)  Z(mm)

-------------------------------------------------------------------------------

4        94   167.08 6.113  0.000  0.685  0.000  -55.16 -22.43 51.12 

3        74   128.65 6.505  0.000  0.517  0.000  -44.74 -31.35 44.44 

1        34   50.36  4.586  0.000  1.000  0.010  -31.88 -53.18 63.40 

5        13   28.02  5.911  0.000  0.771  0.284  -43.57 17.30  43.01 

2        16   17.44  7.078  0.000  0.315  0.834  -29.51 -45.02 66.19 

-------------------------------------------------------------------------------

Height threshold: T = 3.61, p = 0.002 (1.000)

Extent threshold: k = 17.00 sqmm, p = 0.067 (0.854)

Expected sqmm per cluster, <k> = 6.298

Expected number of clusters, <c> = 1.92

Expected surface above threshold (sqmm) = 180.02

Degrees of freedom = [1.0, 12.0]

Smoothness FWHM = 6.2 (mm)

Search vol: 100582 sqmm; 2619.7 resels

=======================================================

 

 

Left: A t-statistics calculated on the cortical surface of the PALS-Atlas (van Essen, in press).
Right: A statistical table generated on the same statistical map using caret_list.

 

A detailed description of the approach taken here and the functions can be found in the following pdf-document:

pdf format Diedrichsen, J. (2005). Surface Statistics using Caret. Resource available online at http://www.bme.jhu.edu/~jdiedric/download/Caret_surface_statistics.pdf.

Download the toolbox
, unzip the archive, and place the folder "caret" in your matlab-path.
Example data for the tutorial is available here.
The toolbox provides the following functions:

1. File-IO
  caret_load  loads a metric,coord,topo,paint,surface_shape into a structure
  caret_save  saves a caret-structure into a file
  caret_savemetric  saves a Metric-structure as file
  caret_savepaint  saves a Paint-structure as file
  caret_struct  generates a new structure to hold a file
2. Surface operations
  caret_getsurface  loads a .coord and .topo file into a surface structure
  caret_calcarea  calculates the Euler Characteristic, circumference and area of the search region
  caret_calcneighbor  makes a list of neighborhood relationships (needed for caret_cluster)
  caret_surfacesubset  makes a subset of surface for small-volume correction
3. Metric file operations
  caret_regionstat  Extract the value from metric files based on ROIs specified in a paint file
  caret_mask  Masks out areas from a metric file, by applying logical mask, specified in a different metric file
  caret_threhold  thresholds a metric file on a height and clustersize threshold
4. Statistical Maps
  caret_getcSPM  makes a statistical map from metric data (using a GLM)
  caret_savecSPM  saves a statistical map as a metric file
  caret_estsmoothness  estimates the smoothness of a statistical map
  caret_clusters  finds clusters on statistical map
  caret_P  calculates corrected, and cluster-wise p-values
  caret_list  lists maxima, clusters, p-values