Dear list and especially Martin,
we have used Freesurfers crosssectional as well as the longitudinal stream and are inspecting the outcomes at the moment.
We have as quality assurance been inspecting regions where the difference between initial CROSS results and the corresponding LONG results exceeds 10 %. In this process we found that for certain regions it seems that the volume increases in almost every case. Below is an overview of differences between cross and long with mean, sd, max and min values in percent for different regions.
We also manually checked an example of the Caudate where there was an increase of 14 % from CROSS to LONG. This difference can be seen between CROSS to LONG for time point 2. It seems like the longitudinal is correct when it covers more of the Caudate. The other time point for that subject and their base seems to provide extra information which is a good thing.
However our question is why this seems to be a systematic thing that all scans for the caudate region increases in volume when doing the longitudinal processing? This is also the case for some other regions as seen below.
Kind regards/
Melanie and Mark
---------------------------------------------------------------------------------- Percent volume changes between CROSS and LONG over 110 subjects: Region Mean Sd Max Min Estimated brain Volume 1 4 32 -18 Total Gray Matter Volume 1 1 5 -2 White Matter Volume -1 1 2 -5 Caudate 5 2 14 0 Putamen 8 3 22 1 Thalamus 2 3 13 -10 Amygdala -1 4 33 -8 Hippocampus 1 4 33 -5 Ventricles 0 2 5 -5 Frontal 0 2 6 -7 Parietal 2 2 8 -6 Temporal 1 2 5 -5 Occipital 1 2 9 -5
Hi Melanie,
whenever you change anything consistently in the processing pipeline (extra steps) you can expect a consistent change in the results. The longitudinal pipeline goes through many different steps than the cross-sectional. E.g. for surfaces, they get initialized from the subject template and then fine tuned, this leads to a consistent shift. Similarly for subcortical volumes, we use a fusion for the aseg, which introduces consistent differences.
Generally these differences should be small and (on average) they should improve reliability and sensitivity of the methods.
Best, Martin
On 05/01/2017 10:17 AM, Melanie Ganz wrote:
Dear list and especially Martin,
we have used Freesurfers crosssectional as well as the longitudinal stream and are inspecting the outcomes at the moment.
We have as quality assurance been inspecting regions where the difference between initial CROSS results and the corresponding LONG results exceeds 10 %. In this process we found that for certain regions it seems that the volume increases in almost every case. Below is an overview of differences between cross and long with mean, sd, max and min values in percent for different regions.
We also manually checked an example of the Caudate where there was an increase of 14 % from CROSS to LONG. This difference can be seen between CROSS to LONG for time point 2. It seems like the longitudinal is correct when it covers more of the Caudate. The other time point for that subject and their base seems to provide extra information which is a good thing.
However our question is why this seems to be a systematic thing that all scans for the caudate region increases in volume when doing the longitudinal processing? This is also the case for some other regions as seen below.
Kind regards/
Melanie and Mark
Percent volume changes between CROSS and LONG over 110 subjects: Region Mean Sd Max Min Estimated brain Volume 1 4 32 -18 Total Gray Matter Volume 1 1 5 -2 White Matter Volume -1 1 2 -5 Caudate 5 2 14 0 Putamen 8 3 22 1 Thalamus 2 3 13 -10 Amygdala -1 4 33 -8 Hippocampus 1 4 33 -5 Ventricles 0 2 5 -5 Frontal 0 2 6 -7 Parietal 2 2 8 -6 Temporal 1 2 5 -5 Occipital 1 2 9 -5
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Martin Reuter -
Melanie Ganz