mri_surfcluster - This tool allows you to cluster surface data.


This program assigns each vertex on a cortical surface to a cluster based on the distribution of intensity values in the source file. A vertex must meet two criteria to be part of a cluster. First, its intensity must fall within a certain range (the intensity threshold criteria). Second, it must be part of a contiguous set of vertices that meet the threshold criteria and whose surface area is greater than a minimum.

There are three types of output. (1) Summary file - simple text file with a list of clusters found including maximum value, area, and talairach coordinate. (2) Filtered - identical to the input except that vertices not assigned to a cluster are set to 0. (3) Cluster number surface - this has the same format as the input and filtered output except that the value at each vertex is the cluster number assigned to it.


--in inputfile

This is the input data to the clustering program. Reads all formats supported by mri_convert.

--subject subjectid

Surface values are defined on this subject.

--hemi hemi

Specify the cortical hemisphere that the input represents. Valid values are lh and rh.

--surf surface

This is the surface to use when computing the talairach coordinagtes. Default is white.

--annot annotationname

Report the cortical annotation that the maximum in each label falls into. The annotation used will be SUBJECTS_DIR/subject/label/hemi.annotationname.annot. Eg, --annot aparc

--frame frameno

Zero-based frame number of the input file. Default is 0. For paint format, zero is the only possible value.

--thmin minthresh

Minimum threshold in the intensity threshold criteria. Vertices on the surface are excluded or included, in part, based on their intensity. For a vertex to be included, its value (ie, intensity) must be within a certain range (namely, between the minimum threshold, set with --thmin, and the maximum threshold, set with --thmax). While thmin and thmax must be positive values, the sign of the threshold can be set with --thsign. However, if thmax is negative, then the maximum threshold will be set to infinity. For example, if --thmin 2 --thmax 5 --thsign pos, then all vertices with values between (positive) 2 and 5 will be candidates for clustering. However, if --thsign abs is used instead, then all vertices between -2 and -5 as well as between +2 and +5 will be candidates.

--thmax maxthresh

Maximum threshold in the intensity threshold criteria. If negative, then the maximum threshold is infinity. See SETTING THE CLUSTER INTENSITY THRESHOLD CRITERIA below.

--sign threshold sign

This is used to control the sign of the threshold criteria. Legal values are pos, neg, and abs. See SETTING THE CLUSTER INTENSITY THRESHOLD CRITERIA below.


Do not adjust thresh for one-tailed tests. By default, the threshold will be adjusted when the --sign is pos or neg by subtracting log10(2.0). This assumes several things: (1) the source is a -log10(p) map, and (2) the pvalue was computed using a two-sided t-test. Under these conditions, subtracting log10(2.0) is like dividing the p-value by 2 (ie, changing it from a two-tailed test to a one-tailed test). If the input map does not meet these criteria, then run with --no-adjust.

--fdr FDR

Set thmin with False Discovery Rate. 0 < FDR < 1. Usually something like .01 or .05. Only when input is -log10(p).

--csd csdfile <--csd csdfile>

Load one or more CSD files. CSD stands for 'Cluster Simulation Data'. This file is produced by running mri_glmfit with --sim. The the threshold and hemi info are obtained from the CSD file and cannot be specified on the command- line. If more than one CSD is specified, they are merged into one CSD internally. When a CSD file is specified, three more columns are added to the summary table:

  1. CWP - cluster-wise pvalue. The pvalue of the cluster corrected for
    • multiple comparisons
  2. CWPLow - lower 90% confidence limit of CWP based on binomial distribution
  3. CWPHi - upper 90% confidence limit of CWP based on binomial distribution

--cwsig cluster-wise_map

Saves the sigificance map of the clustersthe clusters in which the value of each vertex is the -log10(pvalue) of cluster to which the vertex belongs (the cluster-wise significance). The user can also specify that the vertex-wise significance be computed and saved with --vwsig. The significance is based on the distribution of the maximum significances found during the CSD simulation.

--csdpdf csdpdfile

Compute PDF/CDF of CSD data and save in csdpdffile. This is mostly good for debugging.


Only write the csd pdf file.

--minarea area

Minimum surface area (in mm^2) that a set of contiguous vertices must achieve in order to be considered a cluster.

--clabel labelfile

Constrain cluster search to be inside or outside clabel. By default, it will be constrained to be INSIDE. For OUTSIDE, specify --clabelinv. clabel must be a surface-based label.


Constrain cluster search to be OUTSIDE mask or clabel.

--sum summaryfile

The summary text file (the argument of the --sum flag) will contain a summary of the result of the clustering as well as a summary of the conditions under which the clustering was performed. It will list the clusters (1 to N) along with the maximum value found in the cluster (Max), the vertex at which this maximum value was found (VtxMax), the surface area of the cluster (Size), and the Talaiarach coordinates of the maximum (based on talairach.xfm).

--o outputid

File in which to store the surface values after setting the non-cluster vertices to zero. Fmt is the format (currently, only paint format is supported). Note: make sure to put a ./ in front of outputid or else it will attempt to write the output to the subject's surf directory.

--ocn ocnid

File in which to store the cluster number of each vertex. This can be useful for determining to which cluster a particular vertex belongs. It can be viewed as with any other surface value file. Fmt is the format (currently, only paint format is supported). Note: make sure to put a ./ in front of outputid or else it will attempt to write the output to the subject's surf directory.

--olab outlabelbase

Save output clusters as labels. There will be a label file for each cluster. The name will be outlabelbase-XXXX.label, where XXXX is the 4-digit, zero-paded cluster number. The stat field in the label will be the value of the input statistic at that vertex.

--xfm xfmfile

This is a transform file that is used to compute the Talairach coordinates of a vertex for the summary file. The file must be found in subjects_dir/subjectid/transforms. The default is talairach.xfm which is based on the MNI atlas (see --fixmni).

--fixmni --nofixmni

Fix (or do not fix) MNI Talairach coordinates based on Matthew Brett's transform. See Default is to fix. The user may elect not to fix if the xfm file is not talairach.xfm (see --xfm).

--sd subjects_dir

This allows the user to change the FreeSurfer subjects's directory from the command-line. If unspecified, it will be set to the environment variable SUBJECTS_DIR

mri_surfcluster (last edited 2008-04-29 11:46:20 by localhost)