The next step in image preprocessing involved aligning the scans along the anterior and posterior commissures. Such an alignment is necessary due to the varied positioning of the subjects within the magnetic field of the scanner. The program utilized, acpcdetect, is part of the Automatic Registration Toolbox program (NITRC) and was downloaded from the following URL:
https://www.nitrc.org/frs/?group_id=90. The version of acpcdetect utilized was 2011-04-05. The code utilized was as follows:
art/acpcdetect -M -o $subjDir/t1/acpc.nii -i $subjDir/t1/t1.nii
The -M option commands the program to make the midpoint between AC and PC the center of the output field of view.
After aligning the images along the x-,y-, and z- planes in this manner, it was necessary to correct the bias field and remove irregularities in contrast and shading within each individual image. Along with making the image visually more uniform, this process also removes error when applying tissue segmentation techniques to the images. The program Advanced Normalization Tools (ANTs) (STNAVA) (STNAVA) was used for this process and was downloaded from the following URL:
http://stnava.github.io/ANTs/. The version of ANTs utilized for the study was ANTS Version 2.2.0.dev1-g8479e and the code was as follows:
ants/bin/N4BiasFieldCorrection -d 3 -i %subjDir/t1/acpc.nii \
-o $subjDir/t1/n4.nii.gz -s 4 -b [200] -c [50x50x50x50,0.000001]
The -d flag signifies that the input image was three dimensional, the -s flag is a shrink factor of 4 which is the most commonly accepted factor as part of the preprocessing process. The -b flag is bspline-fitting and 200 is also a commonly accepted quantity for such a study. The final flag -c is the convergence or the number of iterations run by the program, 50 being a commonly used value.
After this step was completed, the final remaining step in image preprocessing was to resample all images to have a voxel volume of 1 mm isotropic. This was done using the software Convert3D (ITKsnap) which was downloaded from the URL:
http://www.itksnap.org/pmwiki/pmwiki.php?n=Downloads.C3D. The version of Convert3D utilized in this study was 1.1.0. The program was run with the following code:
c3d $subjDir/t1/n4.niingz -resample-mm 1x1x1mm \
-o $subjDir/t1/resampled.nii.gz
In order to facilitate the preprocessing process, these four steps were combined into a script that contained all of the codes in order. This in turn was added to a secondary script that ran all participant images through the preprocessing process.
Measuring Length of Third Ventricle
After preprocessing the MRI images of both patient and control groups, the NiFTi files were downloaded onto a local computer and analyzed with Mango software (Research Imaging Institute, UTHSCSA). On Mango, the trace function was utilized to measure the lengths of the Third Ventricles in both groups by tracing a line from the back edge of the Anterior Commissure to the frontmost edge of the Posterior Commissure. This process was done manually which could result in minor human error but which should be insignificant compared to the overall measurements collected.
Measuring Volume of Individual Structures in the Ventricular System
The program Freesurfer (Harvard University, Massachusetts General Hospital) was utilized to determine the volume of the right and left hemisphere lateral ventricles, as well as third and fourth ventricle volumes in all patients. The program was downloaded from the following URL:
https://surfer.nmr.mgh.harvard.edu/fswiki/DownloadAndInstall. The program version utilized for the analyses was Freesurfer 6.0.0. The full suite of Freesurfer programs was used with the recon-all function which includes running a variety of computations on the inputted NifTI file including skull stripping, subcortical segmentation, white matter segmentation, cortical parcellation, among many others. It was necessary to run the full suite due to the nature of the program in running the desired subcortical segmentation later in the process. The code input was as follows: