External Email - Use Caution
Hi Maksim,
1. Yes the median image will be very similar to the first three time points. In fact, if the 4th image was an MRI of a knee it would not matter at all. 2. There is a potential bias this way, however, I expect it to be very small. I don’t know of anyone ever investigating this thoroughly.
A potential workaround: Just create the template with the first and the last time points only by passing them only in the base creation step. Then add the other time points into the existing base later, I think there was an add_tp script for that. The disadvantage of this approach is that the base is the mean image between to time points that are further apart, leading to blurry edges in the image, and less clearly defined surfaces. So you remove bias, but increase the chance of reduced quality surfaces slightly.
My recommendation: in a catch-22 I would choose the easier option, which is include all in the base and hope the bias is small. 2 years are not that much different, unless there is really heavy atrophy. We are talking about adult brains, right?
Best, Martin
On 26. Sep 2023, at 13:34, Maksim Slivka maksim.slivka@psykologi.uio.no wrote:
Hi FreeSurfer Community,
I plan to use longitudinal FreeSurfer to process a set of longitudinal scans that are likely to create a bias in the base template. Specifically, our study consists of 3 scans acquired within the first 20 weeks and a subsequent scan acquired 2 years after the third scan.
From the original publication on longitudinal FreeSurfer, I understand that the longitudinal scans are treated equally and do not require to be evenly distributed in time, and that the cross and base images are only used to initialize the longitudinal processing. However, my concern is that the resulting template, based on the median, will be too similar to the first three scans.
Would this negatively impact the results of the 4th scan, due to the inherent bias of the template? Has any group investigated this before?
Any feedback would be appreciated!
Best regards, Maksim
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edumailto:Freesurfer@nmr.mgh.harvard.edu https://secure-web.cisco.com/1p4peD8ajF7StNbsMcsW_4ysU5_UQiUir9y2SziaWTSX8HB...