Hello,
MR spectroscopy was taken with a depth of 7 slices along the SI axis using a GE scanner. The region of interest from which the MRS was taken is manipulated by hand using a mouse with GE software. It appears that LPS (right-handed) coordinate system or the RAS (right-handed) coordinate system cannot be specified when placing the MRS region on the scan. Also, the MRS voxel size is different from the T1 voxel size obtained during the scan.
After the scan, Dicom images were converted to Nifti format with dcm2nii. The Pfile from the scanner prints the center slice number in both Z mm (from XYZ) as 66 and S (from RAS) as 48.3. I enter 0 mm for both the X and Y axes and 0 for the Scanner Anatomical coordinates for both R and A axes. When entering both Scanner Anatomical (0,0,48.3) and XYZ (0,0,66) coordinates into FSLview the printed center slice number in the Pfile is correct for the Scanner Anatomical coordinate and one slice too superior for the XYZ coordinate. We assume that this is due to FSLview including 0 as the first slice?
We verify the correct location of the center slice by looking at a jpg screen shot of the GE monitor when the MRS was done. The center slice number on the screen shot corresponds to the Pfile XYZ coordinate (0,0,66); however, this number, 66, is given in the Pfile as the slice where the MRS was prescribed. No specific coordinates are shown in the jpg of the scanner screen shot. We are certain that the correlations between the Pfile, the screen shot, and the Nifti images we have allow us to select the correct Nifti image to view in FSLview.
This screen shot gives a 2D graphic of the center and borders of the MRS region of interest in sag, cor, and axial views, so the position of the MRS roi is roughly known. The Pfile reports the coordinates of the MRS roi in an RAS coordinate system. It seems that the scanner interprets the MRS box made with a mouse using an RAS system. I assume that transformation matrices are used to generate the RAS coordinates for the MRS roi, but I can't seem to find them.
The RX xyz csi volume dimensions from the Pfile appear to correlate to the mm volume of the MRS roi, which was made standard across subjects, but the reported x volume should be the y volume (and the y volume should be the x volume) . When entering the RX xyz csi volume center into FSLview, the center is not in the correct position (even if we swap the x and y values). I don't know if the RX xyz csi values correspond to the center of the MRS roi, the center of the MRS's center voxel, or the center voxel of the T1 scan. I assume it's MRS related because the volume reported correlates to the MRS roi.
The Pfile also reports the (R,A,S) Center coord of plane image. When these coordinates are entered into the Scanner Anatomical options in FSLview, the center is correct but it is one slice too inferior. This is confusing because when entering (0,0,48.3) the correct slice is viewed along the SI axis (axial plane).
The Pfile gives the coordinates of the top left, top right and bottom right handed corners. Do these values correlate to the MRS center voxel or the MRS roi or the center voxel of the T1 scan? When I enter them into the Scanner Anatomical spaces in FSLview the R, A, and S coordinates are off for all considerations. It seems that when entering the R and A values into Scanner Anatomical values in FSLview they need to be swapped along with a sign change to the R value? But this is uncertain. It is something we have just intuited at best. I assume that transformation matrices converting the MRS roi coordinates to the LPS Dicom format to the RAS Nifti format are not printed in the Pfile, Dicom header, or Nifti header.
I have viewed the sform matrix and the offsets using fslhd -x, calculated the dicom matrix by using cross products of the two-column two-row array with correct sign changes, and the inverses of both of these.
The MRS roi cannot be exported after making it with the GE scanner software.
Our goal is to reproduce the MRS roi in FSLview so that coordinates may be used to extract the exact roi using fslroi. Right now we have a rough roi at best using the screen shot. Metabolite data will then be generated from a proton density roi of the same and location as the MRS roi. The PD roi is flirted to the space where the MRS was prescribed. Resampling is done using a variety of fsl tools.
I read on this forum that FSLview has problems reading "radiological data" because it "prefers" "neurological data." I don't know which of these two formats the data is collected from the GE scanner. I also tried using functions like fslorient and fslswapdim to try to remedy the problem of using the Pfile RAS coordinates to reconstruct the MRS box or the MRS voxel in FSLview.
Could someone please help? I realize that there are numerous issues raised here and that each requires a fair amount of depth.
Thank you,
Alex
|