My off the cuff suggestions would be potentially signal to noise issues. If the Timm gave the highest SNR, this could result in higher FA values across much of the cortex.
If the samples are smaller, it could also be blind luck. Or there could be a recruitment bias (e.g. the Timm site recruited a bunch of grad students, the other site recruited in the community and had a higher proportion of lower-functioning individuals with lower FA).
Sorry I have no advice on how to check SNR in DTI data. I have personally only done a bit of DTI work.
Best of luck,
Colin Hawco, PhD
Neuranalysis Consulting
Neuroimaging analysis and consultation
www.neuranalysis.com
[log in to unmask]
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Jack Rogers
Sent: October-18-16 9:00 AM
To: [log in to unmask]
Subject: [FSL] Multi-Site DTI Data Analysis/TBSS
Dear Experts,
I am currently in the process of analysing DTI data acquired across 4 sites (N=300+ individuals). The sequences are comparable and QA checking was carried out prior to data collection.
In addition to including site as a covariate I have also modelled each site explicitly as an EV to compare mean FA values between sites. I notice with some concern that one site reveals significantly increased FA (TFCE, p<.05 FWE-corrected) across (almost) the entire FA skeleton when compared to the remaining three sites. Could anyone advise on why this might be and checks I could carry out to get to the root cause of this. If it helps, the site in question used a Siemens 3T Tim Trio whilst the remaining sites used a Siemens 3T Prisma and we (Birmingham) used a 3T Phillips Achieva. All other sites (i.e. 2X Siemens Prisma versus Phillips Achieva) seem equivalent.
Many thanks for you help in advance.
Jack Rogers (Post-doctoral researcher at the University of Birmingham)
|