With the Caret Map fMRI to Surface program (map_fmri_to_surface), you can map functional data either onto the Human Colin atlas surfaces or onto surface reconstructions of individual brains (human, monkey, or other species):
The grids supported by the Caret mapper are among the many standard stereotaxic spaces (e.g., Talairach, MNI152) to which neuroimaging data can be mapped. They differ in overall volume dimensions, voxel sizes, and center. As we become aware of other popular versions, we will add mapper support for them. Email caret@v1.wustl.edu to nominate candidates for mapper atlas support.
Caret can map the following types of statistical output volumes onto the Human Colin atlas:
For each grid, the dimensions and location of the origin* are listed in the following table.
| Grid name | Voxel Dimensions | Volume Dimensions | Origin* |
|---|---|---|---|
| 711-2B | 1x1x1mm | 176x208x176 | 88,122,73 |
| 2x2x2mm | 128x128x75 | 64,63,33 | |
| 3x3x3mm | 48x64x48 | 23,34,20 | |
| AFNI | 1x1x1mm | 161x191x151 | -80,-80,-65** |
| SPM99 default bounding box 78:78 -112:76 -50:85 | 1x1x1mm | 157x189x136 | 78,112,50 |
| 2x2x2mm | 79x95x68 | 39,56,25 | |
| 3x3x3mm | 53x63x46 | 26,37,17 | |
| SPM99 template bounding box -90:90 -126:90 -72:108 | 1x1x1mm | 182x218x182 | 90,126,72 |
| 2x2x2mm | 91x109x91 | 45,63,36 | |
| 3x3x3mm | 61x73x61 | 30,42,24 | |
| SPM2 default bounding box 78:78 -112:76 -50:85 | 1x1x1mm | 157x189x136 | 78,112,50 |
| 2x2x2mm | 79x95x69 | 39,56,25 | |
| 3x3x3mm | 53x63x46 | 26,37,17 | |
| SPM2 template bounding box -90:90 -126:90 -72:108 | 1x1x1mm | 181x217x181 | 90,126,72 |
| 2x2x2mm | 91x109x91 | 45,63,36 | |
| 3x3x3mm | 61x73x61 | 30,42,24 | |
| * The origin is roughly the anterior commissure (AC). For 711-2B and AFNI, the origin is the intersection of the AC-PC line and the posterior AC margin. For SPM users, the origin is the SPM originator coordinate less one along each axis, because SureFit/Caret voxel 0,0,0 is 1,1,1 in SPM. This voxel is actually above the AC; see Matthew Brett's "The MNI brain and the Talairach atlas". | |||
| ** The AFNI origin is expressed as the value returned by 3dAttribute ORIGIN; this volume is in "RAI" orientation (right-to-left, anterior-to-posterior, inferior-to-superior). AC coordinates for all other grids are given in "LPI" orientation (left-to-right, posterior-to-anterior, inferior-to-superior). | |||
You must spatially normalize your functional data in SPM before mapping it to the atlas. Functional data in "magnet space" is not likely to align properly with the surface grids. (This comment applies to the 711-2B and AFNI spaces as well; i.e., those volumes must be registered to 711-2B or +tlrc, respectively.)
In spm2, spatially normalized output can be in radiological (right-to-left) or neurological (left-to-right) orientation. In previous versions, we could assume spatially normalized volumes were oriented left-to-right (neurological). The SPM surfaces used for mapping are all in neurological convention, so to map accurately, labs using radiological normalized output (i.e., analyze.flip=1 in spm_defaults.m) must specify a flip when mapping to the atlas. See the Functional Imaging Laboratory's (FIL) SPM2 page, Compatability section for further details.
The volumes produced by spatial normalization under these two versions looked sufficiently different to merit separate mapping surfaces. If you want the best fitting surface for mapping your functional data, you should choose the surface corresponding to the SPM version used to spatially normalize your data. If your functional data was normalized using a version of SPM prior to spm99 (e.g., spm96b), then map to the spm99 surfaces. If you want to compare the surfaces, load the corresponding FIDUCIAL coord files in Caret and see the differences for yourself. The SPM/AFNI volumes also are provided, so you can compare them in SPM, AFNI, or Caret5 (yes, Caret can display volume files now!). These volumes should all be in MNI152, and provided that your volume has a valid originator voxel coordinate, the mapper should behave well.
Since too many ventral nodes are truncated (numerous "NaN NaN NaN" -- "not a number" -- coordinates returned by norm_coord.m -- see SPM2 Spatial Normalization of Human Colin Atlas Surfaces), we were not able to generate a cerebellum surface for the default bounding box. Of course, if you are studying the cerebellum, you probably are using the template bounding box anyway. If not, there is a cerebellum surface for the default bounding box generated using spm99, but much of it is truncated.
The mapper cares whether you wrote your volume in default or template bounding box. Make sure you choose the corresponding surface, based on the bounding box used for writing normalized in SPM.
The procedures used to normalize the "native" colin surfaces are described here:
Caret needs the following input data for mapping functional data onto an individual's surface:
If your anatomical and functional volumes have been spatially normalized in SPM, then you just need to write your normalized functional volumes in the same bounding box (default or template) and voxel dimensions (cubic 1mm for humans) as your anatomical volume.
You may need to coregister your functional volumes to your anatomical volume. If your anatomical volume was acquired in sagittal or coronal slices, then it needed axes permutation before processing in SureFit. In this case, you will need to use SPM's Display utility and find the proper pitch, roll, and yaw parameters to make the images in SPM's Display window appear in the proper orientation, as shown in the image below. For example, pitch=pi/2; roll=pi/2; yaw=0; resize{x}=1; resize[y]=1; resize[z]=-1 will reorient sagittal slices (ASR orientation) to axial. Pressing Reorient Images will write out a .mat file, which can be used to reslice the functional volumes using SPM's Co-register menu.
Using graphical interface:
AFNI's adwarp command can be used on the command line or in a shell script, if you want to streamline preprocessing for multiple volumes.
The current mapper is part of Caret; see the Caret Home Page.
For most users of the stand-alone mapper, the Tutorial 1 dataset is probably the easiest way to download the atlas. More current versions of the colin left, right, and cerebellar surfaces are available on the SuMS home page. But the stand-alone mapper also needs the MAP_MASTER_FILES subdirectory, which contains a file named map_fmri_atlas.directory. This file cites the target spec filename, which now is target space dependent in the SuMS datasets (e.g., SPM, AFNI, or 711-2B). The tutorial 1 dataset contains all three colin surfaces and the MAP_MASTER_FILES subdirectory, but is updated less often than the SuMS datasets. John Harwell is working on a version of the mapper that will support easier updates.
There are separate tutorials for atlas and individual mapping:
SureFit retains some menu-driven and command line mapping features for long-time users who still find them useful. But the Caret mapper is superior in many ways (e.g., supports more data types; supports atlas mapping; can read multi-bucket AFNI .BRIKs). SureFit's menu-driven mapper only supports 16bit short volumes. Its command line utility Map2Surface supports more input data types and map methods (e.g., avgvoxel), but it still lacks many of the Caret mappers recent enhancements. Some wustl/MCW users may still use the Map2Surface.sh shell script, which calls the Map2Surface command line utility. As long as you find these tools useful, feel free to use them. Just keep in mind that any fixes/enhancements will be made to the Caret mapper -- not SureFit's mappers.
Some users may not be aware that there is a command line interface to the Caret mapper. Enter 'map_fmri_to_surface -h' at the command line for usage.
For comments or more info, contact webmaster@v1.wustl.edu.
© Van Essen Lab; all rights reserved.