WM Atlas and the PAM50_cord do not match in native space

Dear experts,

When I warp the template and all atlases to a native space image using the sct_warp_template script, I found that the atlas and the template PAM50_cord do not match exactly. What could be the reason?

Scripts that I used:

sct_register_multimodal -i ../XX-T2_SAG_resize_seg_labeled.nii -d XX_eddy_md_crop.nii -x nn -identity 1
sct_label_utils -i XX-T2_SAG_resize_seg_labeled_reg.nii -vert-body 1,10 -o XX_dmri_labels.nii.gz
sct_register_to_template -i XX_eddy_md_crop.nii -l XX_dmri_labels.nii.gz -s /home/jnavas/SPINALCORD/spine_masks/masks_reg_crop_fa/XX_WholeSpine_Mask2_reg.nii.gz -c t1 -param step=1,type=seg,algo=centermass:step=2,type=seg,algo=bsplinesyn,metric=MeanSquares,smooth=1,iter=6
sct_warp_template -d XX_eddy_md_crop.nii -w warp_template2anat.nii.gz -a 1 -s 1 -ofolder label

Hi @Javier_Navas,

The cord mask (PAM50_cord.nii.gz) is binary while the PAM50 white/gray matter atlas is soft/probabilistic (ie: float between 0 and 1). Details about the generation of the WM atlas can be found in Levy et al. Neuroimage 2015.

So, depending how the atlas is thresholded on the visualisation, there will be a variable amount of overlap between the cord mask and the thresholded soft PAM50 atlas mask. See illustration below, where the threshold varies from 0 to 0.9 on the WM atlas:

Hope that answers the questions.
Julien