 |
 |
If you register image 1 to image 2, rather than what you just did, do you find the same difference? If not, the problem might be that the registered image has been resampled in the registration process and this somehow influenced the FA values. As an aside, if you are doing a registration in the same subject you should only be using 6 dof. Also, have you measured the SNR of the images and found them to be the same? Lower SNR tends to result in increased FA values. Finally, do you know that there have been no coil changes, scanner software or hardware upgrades in the intervening time? Peace, Matt. -----Original Message----- From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Walterfang, Mark Sent: Sunday, January 04, 2009 7:02 AM To: [log in to unmask] Subject: [FSL] Longitudinal DTI Hi all Long time lurker, second time poster. My question relates to longitudinal DTI measurements in subjects to track illness progression and/or response to a novel treatment. My subjects are two adults who were scanned approximately 12 months apart using the same 15-direction sequence on a GE Signa 1.5T scanner. The parameters were identical between times 1 & 2 with regards to the sequence itself (TE, slice thickness, etc). On each of the scans I converted from DICOM to NIFTI using Jolinda's wonderful MRIConvert, and then ran eddy_correct, extracted the nodif volume, used BET on the nodif, and then ran dtifit to produce the FA and MD maps I am interested in. This seemed to work fine for all scans. I then used FLIRT with 12 dof to register the time 2 FA map to the time 1 map, and they match nicely from an anatomical perspective in FSLView. I am looking to create ROIs in FSLView to compare values in particular regions across timepoints. What has occurred is that the FA values in one patient have dramatically reduced over 12 months, in a counter-intuitive way (given clinical progress & response to medication) - with reductions varying from 0.07 to 0.15 in all regions, brain-wide. The other patient, with a virtually identical clinical response, has shown an increase that is fairly consistent across regions of 0.05. What I am wondering is essentially how valid these changes are. Could processes such as eddy current correction, or other scanner/processing variables, produce brain-wide changes in the FA maps? Are there other pre-processing steps I need to consider to improve (if it's necessary) the robustness of my output FA maps? Any help is appreciated, Regards Mark Walterfang