MCALT version 1.3 Release Posted By: Christopher Schwarz - Sep 24, 2018Tool/Resource: Mayo Clinic Adult Lifespan Template and Atlases MCALT version 1.3 is released! This release adds three new files to MCALT: 1) a 0.5mm T1-weighted template image; 2) a mask of voxels inside the cranial vault, i.e. the TIV; 3) a segmentation mask designed to remove misclassified adjacent regions of gray-matter-isointense dura from medial temporal lobe (MTL) segmentations. * ./MCALT_T1_05mm.nii A higher-resolution (0.5mm) version of MCALT space. This is not the version on which all MCALT atlases were drawn and defined (MCALT_T1.nii). We created this version later, intended primarily for visualizations. It is possible to use this 0.5mm version for atlas propagation, but this has not been rigorously tested. To create this image, the same 202 (~1mm T1-weighted) MRI as in the standard MCALT template were coregistered using buildTemplateParallel.sh from the ANTs package, using a copy of MCALT_T1.nii up-sampled to 0.5mm as a rigid reference target. A single warp (masked to prioritize matching brain voxels) was then calculated between the generated 0.5mm template image and the standard (1.5mm) MCALT_T1.nii. This final warp improved the voxel-wise correspondence between the template versions, thus making it possible to consider this new image to be in the same voxel "space" as MCALT, rather than a new and incompatible template space. All 202 scans were warped via the transforms generated by buildTemplateParallel concatenated with the final warp across templates, then averaged, to create the final MCALT_T1_05mm.nii * ./MCALT_TIVmask.nii A mask of MCALT voxels that exist inside the cranial vault, i.e. the total intracranial volume. * ./atlas/MCALT_MTL_nogo.nii MCALT_MTL_nogo is a mask designed primarily to remove misclassified adjacent regions of gray-matter-isointense dura from medial temporal lobe (MTL) segmentations. It contains non-brain voxels, proximal to the medial temporal regions, that SPM12 segmentations can sometimes misclassify as brain despite their being labelled as non-brain in our MCALT_tpm. For better segmentation accuracy, we recommend that this mask be applied after segmentation to set any contained voxels as non-GM and non-WM. We typically apply this mask to improve segmentation results in native space. It may also be reasonable to apply this mask to segmentations in template space, but we have not tested it for that use. |
| Latest News |
|