[Mristudio-users] Handling of post-LDDMM tensor files

[Christian Bieck]

Salutations,

I finally managed to do a proper and repeatable LDDMM processing, thanks to the exceptionally helpful Anthony Kolasny, but have encountered further problems along the way, about which I'd be very grateful to get some input for.
My main goal is to transform tensor information to MNI space so that I may do a fiber tracking for all subjects using the same ROI.
Here's what I did so far:

1. In DTI Studio: Calculated the DWI image and the 6 tensor files.

2. In Landmarker: 
a.) Opened the JHU-MNI_DWI Template
b.) Loaded the DWI file (Image Size 96x96; Image Slices 60; Slice Orientation Axial; Slice Sequencing Inferior-Superior; FOV 192x192; Slice Thickness 2; Pixel Size 2x2) as the subject
c.) Loaded the six tensor files (Image Size 96x96; Image Slices 60; Slice Orientation Axial; Slice Sequencing Inferior-Superior; FOV 192x192; Slice Thickness 2; Pixel Size 2x2) via the "Open .d file", selecting the "X-component for the "Flip Eigen Vector"
d.) Did a Linear AIR Transform with default values and Affine/Trilinear selected, leaving the parameters of the Image and tensor files equal to the parameters of the Template file (Image Size 181x217; Image Slices 181; FOV 181x217; Slice Thickness 1; Pixel Size 1x1)
e.) Converted all images (Template, DWI and tensors) to byte file format with the "BFW" button
f.) Chose a Single channel Volume LDDMM, where the images were first automatically changed to 182*218*182. I then left all the default parameters and sent the files (Template, subject DWI and tensors) to the server.

3. Shortly thereafter, the processing was finished, but instead of a transformed subject DWI and transformed tensors, the zip file contained only a template.img, target.img,the transformation matrixes hmap.vtk and kimap.vtk and some text files.


4. In Landmarker:
a.) Opened the template.img as the template (Image Size 182x218; Image Slices 182; Slice Orientation Axial; Slice Sequencing Inferior-Superior; FOV 182x218; Slice Thickness 1; Pixel Size 1x1)
b.) Opened the target.img as the subject (Image Size 182x218; Image Slices 182; Slice Orientation Axial; Slice Sequencing Inferior-Superior; FOV 182x218; Slice Thickness 1; Pixel Size 1x1)
c.) As there were no LDDMM processed tensor files, I resampled the original tensor files (that were not Linearly AIR'd, maybe that might be a problem) to the parameters of the target.img/template.img
d.) I opened said resampled tensor files via the "Open .d file" button, with Flip Eigen Vector X-component selected
e. I loaded the LDDMM Matrix (VTK Format) via the Load Transformation Matrix button (Is it important whether the Hmap.vtk or Kimap.vtk gets chosen?)
f.) Saved the Updated_Target.img and the Updated Tensor files

5. In DTISTudio:

a.) Opened the Updated_Target.img
b.) Opened the Updated Tensor files via the "Open .d" button,, selecting the "Get eigen-values/etc." option and specifiying the "DTISTudio" format
c.) Saved the resulting Eigenvector0 and FA images for the Fiber Tracking
d.) As my computer had severe problems handling any ROI operations for the original Image size 182x218; Image Slices 182 files, I resampled the Eigenvector0 and FA files to Image Size 96x96; Image Slices 60 with Landmarker and then did the Fiber Tracking (again, with Flip Eigenvector X-component selected) and ROI operations


Now, my questions are:
1. Is the general procedure sound so far or do any of the steps lead to the important tensor information getting distorted to the point of unusability? I still want to track the optic radiation and want to compare FA values (which might have only slight differences between patients and normals), after all.
2. The LDDMM target.img looks very different from the sent subject DWI image even before applying the LDDMM transformation matrix, leading to the Upated Tensor files looking very different from the Updated_target.img. Is that a problem? Do the Tensor files still get properly transformed to MNI space?


Thank you very much for your time and effort,
Christian Bieck