Hello-
We have discovered that the 3T Siemens Trio running B15 sometimes logs different SeriesInstanceUID's for a single acquisition. For example, for a 3D MP-RAGE scan some slices will be saved with one SeriesInstanceUID, whereas others will use a different SeriesInstanceUID. This appears to occur pretty rarely (maybe ~3% of sessions), however when it does occur it will disrupt SPM8's DICOM-to-nifti conversion. We have found the same behavior on two Trios (one at www.cabiatl.com and the other at bitc.bme.emory.edu/index1.shtml), and determined that the error occurs on the Siemens console, not due to DICOM export. We have reported this error to Siemens. In the meantime, there are a couple solutions. One is to use dcm2nii for image conversion and the other is to modify SPM's code (as shown below). Note that these changes to SPM8 may disrupt conversion of images where there is a legitimate difference in SeriesInstanceUIDs, though in our experience the other DICOM fields are usually sufficient to distinguish images. I want to stress this is not a bug with SPM8 per se, rather a hack to correct for a bug in Siemens systems.... It is pretty easy to detect if you encounter this bug, as for example a 3D T1 image with 170 slices will be saved for example as two images with 99 and 81 slices each.
edit spm_dicom_convert.m by making lines 295,296,297 comments rather than code. I.E. add % to the start of each line so they read:
%if isfield(hdr{i},'SeriesInstanceUID') && isfield(vol{j}{1}, 'SeriesInstanceUID')
% match = match && strcmp(hdr{i}.SeriesInstanceUID,vol{j}{1}.SeriesInstanceUID);
%end;
|