Hi, sorry for the delay
On 4/23/2025 10:43 PM, Dorian Pustina wrote:
External Email - Use Caution
Hi all, I have some questions regarding failed segmentations on
longitudinal data.
We have large datasets of Huntington's disease with data collected
longitudinally for 3-10 years. Freesurfer is our main tool to extract
volumetric scores, particularly in caudate and putamen.
I usually QC Freesurfer segmentations and include the fail/pass status
for other researchers to use. I inspect the segmentation of the
cross-sectional pipeline because I noticed that mis-segmentations are
clearer and the longitudinal pipeline does not seem to correct them.
For example, a displaced putamen segmentation in cross-sectional
pipeline becomes just more fuzzy in the longitudinal pipeline with
more peppered voxels at the edges. So basically, my QC of the
cross-sectional segmentation decides if the timepoint is good or not.
I have three questions:
1. Is my understanding correct that a motion corrupted scan (i.e.,
with motion rings) is going to affect the other timepoints by
compromising the SST? Thus the only solution would be to exclude the
motion corrupt timepoint from FS processing? And in case the
timepoint has been processed with FS, is the best practice to exclude
the entire subject given that the bad scan have compromised the
accuracy of other timepoint segmentations?
What is the SST? Is it the base? Then yes, you should exclude a
corrupted time point from the base as it will affect base results. I
don't necessarily think that you have to exclude all time points from a
subject if you accidentally included a corrupted time point in the base.
The base serves to initialize the analysis for the longitudinal time
point. Corruption that leaks into a base could leak into an uncorrupted
long time point, but it might not. If the long looks ok, then you don't
necessarily need to exclude it.
2. On a different scneario, if a timepoint has a good scan but
Freesurfer's segmentation is simply innacurate or displaced, is this
bad segmentation at the cross-sectional level going to affect the
other timepoints after running the longitudinal pipeline?
The cross segs are fused into the base seg and the base seg is used to
initialize the long seg, so a bad cross seg could leak into a non-bad
long seg.
. This is the most important question: is there a common
recommendation or practice how to deal with failed segmentations in
the longitudinal pipeline? Correcting manually the segmentations is
not an option, so I am asking whether we can just exclude the
failed segmentation timepoint from the analysis or whether we should
exclude the entire subject?
I'm not sure what you are asking here. You can't exclude just the
segmentation from the long analysis. You'd have to exclude the entire
time point.
Thank you for sharing any thoughts. I cannot seem to find information
on this topic on papers or the mailing list, so any tips on how you or
your lab deals with these issues can help.
Dorian
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is
addressed. If you believe this e-mail was sent to you in error and the e-mail
contains patient information, please contact the Mass General Brigham
Compliance HelpLine at https://www.massgeneralbrigham.org/complianceline
<https://www.massgeneralbrigham.org/complianceline> .
Please note that this e-mail is not secure (encrypted). If you do not wish to
continue communication over unencrypted e-mail, please notify the sender of
this message immediately. Continuing to send or respond to e-mail after
receiving this message means you understand and accept this risk and wish to
continue to communicate over unencrypted e-mail.
