Example demo input:
sct_deepseg_sc -i t2.nii.gz -c t2 -qc ~/qc_singleSubj
-
Is the input only restricted to NIfTI images? I have DICOM images, but I will find a converter if necessary.
-
I want my output to be segmented images of the spinal cord in axial view, and I don’t need the sagittal view. However, does the input of sct_deepseg_sc need MRI scans that contain both sagittal and axial views? Originally, I was going to strip the input scans of the sagittal view, but it looks like the input in the demo needs the sagittal view that then generates axial view results.
Should we also remove images from the input MRI scans that don’t have the spinal cord in clear view (i.e. outer irrelevant tissue and fat layers with no spinal cord), or is it fine to leave them in?
I’m a complete beginner to all of this, so if anyone could link beginner-friendly resources for these questions or answer them, thank you!
Hi @aliceyu2,
Thank you for reaching out.
Is the input only restricted to NIfTI images? I have DICOM images, but I will find a converter if necessary.
Yes, the input needs to be NIfTI images. So you can convert them using e.g. dcm2niix, which we recommend.
I want my output to be segmented images of the spinal cord in axial view, and I don’t need the sagittal view. However, does the input of sct_deepseg_sc need MRI scans that contain both sagittal and axial views?
No, it does not. Please note that a segmentation is not done in axial or in sagittal view only. A segmentation is done in all view, as it is a process that consists in labeling each voxel (ie: volume element) as “1” (inside the spinal cord) or “0” (outside the spinal cord, a.k.a background). An MRI volume can be viewed in axial, sagittal, or coronal, but the “viewing” process is independent from the “segmentation” process. So all you need to do, is input your 3D MRI volume in sct_deepseg_sc, and then visualize the output segmentation in any view you like. I recommend you follow the latest SCT course to familiarize yourself with the concept of 3D MRI images and their analyses.
1 Like
@jcohenadad has provided an excellent answer. To add onto his answer, though, I’ll just address one specific part of your reply:
To clarify the 2 differing views you see in the tutorial figures: The “sagittal view” you refer to is just a sample figure in the tutorial, used for demonstrative purposes (it shows screenshots of individual sagittal slices taken from a 3D image viewer). On the other hand, the “axial view” you refer to is a separate “QC report” generated by SCT.
QC reports are basically SCT’s built-in way to visualize the 3D output (similar to other image viewers/visualizers that @jcohenadad refers to). One benefit to using SCT’s QC reports is that our slice-by-slice axial view offers a quick way to visually inspect each axial slice of the segmentation all at once, without needing to manually scan through the slices one by one in 3D. This makes it faster to catch poorly-segmented regions in the output.
Kind regards,
Joshua
1 Like