Hi Carissa - Any QA issues aside, the first reason for a difference that I can think of is that the same % threshold for the tract probability distribution (e.g., 20% of the max) would not be equivalent between cross and long streams. In the long stream I'd expect the distributions to be tighter because they're derived from twice as much data, so there is reduced uncertainty.
In the 2016 longitudinal TRACULA paper we looked at test-retest reliability, but that was reliability between two scans of the same subject when analyzed with the same stream. That was higher when the two scans were analyzed jointly with the long stream than when they were analyzed as independent data points with the cross stream. But, there is no expectation of reliability between the cross and the long stream on the same data set. They are two different analysis methods and there is no reason to expect that they will give the same result or that the result will scale linearly. If they did, both streams would give the same answer when comparing two populations, so there would be no reason to have a long stream.
Hope this helps, Anastasia.
________________________________ From: freesurfer-bounces@nmr.mgh.harvard.edu freesurfer-bounces@nmr.mgh.harvard.edu on behalf of Carissa Nicole Weis cnweis@uwm.edu Sent: Thursday, March 25, 2021 6:06 PM To: freesurfer@nmr.mgh.harvard.edu freesurfer@nmr.mgh.harvard.edu Subject: [Freesurfer] cross sectional vs longitudinal TRACULA reliability
External Email - Use Caution
External Email - Use Caution
Hello,
I’ve run both longitudinal and cross-sectional TRACULA pipelines on the same dataset, and I’ve found surprisingly low correlations of FA measurements for several tracts (as low as r= .1 in some cases). Which measurement should be considered more reliable? I’m having a hard time deciding which set of results to carry forward to group analysis.
Thanks,
Carissa