Clinical fMRI ROI masks (ClinROIs)
==================================

These anatomical image files are used for quantitative assessments of BOLD brain activation in clinical fMRI scans.
The images are in compressed NIFTI file format with an accompanying XML header file (BXH -- BIAC XML header).  
The images are in MNI_152 anatomical space -- a copy of the MNI_152 brain images (from FSL) are also included.
 
The XML headers provide the same information as the NIFTI file headers, plus additional information including pointers to
voxel color tables and a glossary of anatomical labels. The XML header syntax was developed at Duke as part of the
NIBIB-funded Functional BIRN project.  It is used by fScan software but can be ignored otherwise.

The .wdf file is a fScan workspace file that describes the anatomical relationships among a group of related image files.
It is also used by fScan software but can be ignored otherwise.

The ClinROI images have different voxel values for different anatomical regions of interest in clinical fMRI.
Of particular interest are the expressive and receptive language areas as well as hand sensorymotor areas.
The mask regions specified in this atlas were defined at Duke. They are originally based on a combination of anatomical ROIs provided in
the AAL and Harvard-Oxford atlases provided in FSL and the Wake Forest Pick-Atlas.  Those maps were then editted based on empirical analysis
of the locations of language and motor activations obtained in over 1000 clinical fMRI scans performed at Duke. In particular, the language
ROIs were adjusted to distinguish anatomical regions that are most often associated with semantic language function (traditional Broca
and Wernicke's areas) from other areas often activated in language fMRI tasks.  The resulting areas were then made left-right symmetrical
in order to allow unbiased assessments of language laterality.

There are 2 sets of ClinROI images. The "ClinROIs" images have different voxel values for ROIs in the left and right hemispheres.
The "ClinROImasks" images have the same voxel values for homologous ROIs in the two hemispheres.  The 2 sets of images facilitate
fMRI threshold adjustments in expressive and receptive language areas, regardless of hemisphere (ClinROImasks), and then comparison
of relative activation strength (e.g. laterality index) between hemispheres (ClinROIs).

The way these images are used at Duke is to first be sure that the patient's fMRI maps are accurately registered to the patient's 
own T1-weighted anatomical images. Once that is done then the atlas MNI152_1mm_T1_brain images need to also be registered to 
the patient's T1-weighted anatomical images.  We do both registration steps using the fScan program but you can use other software
for this.  I know the atlas brain registration can be done using ANTs and I think it can be done using AIR, FSL, SPM, or AFNI. 
For quantitative analysis it is important to register the atlas to the subject, not the subject to the atlas, as measurements 
should be made in real patient brain units.

Once both the atlas and the BOLD activation map are registered to the patient T1 images you can transform ClinROI atlas voxels
directly to BOLD map voxels to identify and measure activations within functionally important ROIs. It is a good idea to first adjust
activation map thresholds in order to only measure voxels with significant activation signals.

At Duke, we set activation thresholds using the AMPLE (activation mapping as percentage of local excitation). For task-relevant ROIs,
the threshold is set independently in each ROI to either a t-value of 4.0 or 50% of the peak activation t-value in that area, 
whichever is greater.  

For motor mapping, thresholds are set independently for right and left hand areas (masks 17 and 49) using just the ClinROIs images, 
and activation is then quantitated using the same images.

For language mapping, thresholds are set independently for receptive and expressive speech areas (masks 1 and 2) using the ClinROImasks
images, and activations are quantitated using the ClinROIs images that distinguish between left and right areas. Laterality indices,
for example, can be calculated separately for receptive language by comparing activations in masks 1 and 33, and for expressive
language by comparing activations in masks 2 and 34.