Dear Paulo,
is this affecting only the sign of the x-dimension or the accuracy of the values? And how large are the differences?
Best,
Christian
On Wed, 19 May 2021 10:35:38 +0200, Paulo R. Dellani <[log in to unmask]> wrote:
>Dear CAT/SPM experts,
>
>I processed a number of MR scans (Siemens 3T MPRAGEs) with CAT 12.7
>(r1733) and noticed that the voxel dimensions of some of the resulting
>NIfTI images in native space, for example mri/p*, differ from the voxel
>dimensions of the NIfTI images with the original scans.
>
>While the differences are very small, software used in the next steps of
>the processing pipeline detect these mismatched voxel sizes and produce
>an error. These differences appear only in the voxel dimension
>corresponding to the left-right axis of the anatomical frame of
>reference, so for scans with slices in sagittal orientation, pixdim3 is
>affected and for axial and coronal orientations, pixdim1. Voxel
>dimensions along the other two axis are not affected. The mismatches are
>100% reproducible.
>
>The original MR scans where converted from DICOM to NIfTI using the last
>version of dcm2niix (v1.0.20210317). CAT segmentation was run using the
>standard parameters. Matlab version is 9.3.
>
>When converting the scans from DICOM to NIfTI using SPM own image format
>conversion tool, the voxel dimensions of the results in native space
>match the dimensions of the original scans. However, since the project
>aims to process a large number of scans (30 000 +) at the moment we are
>not willing to recreate the cohort BIDS.
>
>Any ideas on how to solve this without overwriting the header of the
>affected resulting NIfTI images?
>
>Cheers,
>
>Paulo R. Dellani
>
>
|