I am planning an analysis using freesurfer to compare 2 groups. The SPGRs that will be used vary in regard to slice thickness (from 0.86 to 1.5 in each axis), number of slices acquired (100-256), and orientation (axial; coronal) due to the fact that these were collected over a number of years on the same scanner. In another posting I read that this might be a problem but wanted to revisit this to see if any specific heterogeneity in the parameters should be particularly avoided. Also, if the groups differ such that one group has largely similar scan parameters and the second group is more heterogeneous, will this introduce any systematic bias in cortical thickness estimates, parcellation, or segmentation values? Thanks for any advice on this issue.
Hi James,
it certainly can. If one group had all one scan type and another all a different one then you wouldn't be able to tell if the effects you find are due to the group membership or the scan differences.
cheers, Bruce
On Tue, 6 Oct 2009, James Root wrote:
I am planning an analysis using freesurfer to compare 2 groups. The SPGRs that will be used vary in regard to slice thickness (from 0.86 to 1.5 in each axis), number of slices acquired (100-256), and orientation (axial; coronal) due to the fact that these were collected over a number of years on the same scanner. In another posting I read that this might be a problem but wanted to revisit this to see if any specific heterogeneity in the parameters should be particularly avoided. Also, if the groups differ such that one group has largely similar scan parameters and the second group is more heterogeneous, will this introduce any systematic bias in cortical thickness estimates, parcellation, or segmentation values? Thanks for any advice on this issue.
Does each group have a mix of parameters or does each group have its own set of parameters. If the latter, then it is a real problem.
doug
Bruce Fischl wrote:
Hi James,
it certainly can. If one group had all one scan type and another all a different one then you wouldn't be able to tell if the effects you find are due to the group membership or the scan differences.
cheers, Bruce
On Tue, 6 Oct 2009, James Root wrote:
I am planning an analysis using freesurfer to compare 2 groups. The SPGRs that will be used vary in regard to slice thickness (from 0.86 to 1.5 in each axis), number of slices acquired (100-256), and orientation (axial; coronal) due to the fact that these were collected over a number of years on the same scanner. In another posting I read that this might be a problem but wanted to revisit this to see if any specific heterogeneity in the parameters should be particularly avoided. Also, if the groups differ such that one group has largely similar scan parameters and the second group is more heterogeneous, will this introduce any systematic bias in cortical thickness estimates, parcellation, or segmentation values? Thanks for any advice on this issue.
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
In regard to group differences, roughly 1/3 of the scans in one group match the voxel size and orientation parameters in the other. The remaining scans are uniformly different from each other. Not knowing the specific effects that comparing unequal parameters will have in a Freesurfer analysis, is there a threshold or rule of thumb for when not to compare groups with different parameters? Are there specific parameter differences to avoid (x, y, z, size; orientation)? Thanks again for your help.
On Tue, Oct 6, 2009 at 12:14 PM, Bruce Fischl fischl@nmr.mgh.harvard.eduwrote:
Hi James,
it certainly can. If one group had all one scan type and another all a different one then you wouldn't be able to tell if the effects you find are due to the group membership or the scan differences.
cheers, Bruce
On Tue, 6 Oct 2009, James Root wrote:
I am planning an analysis using freesurfer to compare 2 groups. The SPGRs
that will be used vary in regard to slice thickness (from 0.86 to 1.5 in each axis), number of slices acquired (100-256), and orientation (axial; coronal) due to the fact that these were collected over a number of years on the same scanner. In another posting I read that this might be a problem but wanted to revisit this to see if any specific heterogeneity in the parameters should be particularly avoided. Also, if the groups differ such that one group has largely similar scan parameters and the second group is more heterogeneous, will this introduce any systematic bias in cortical thickness estimates, parcellation, or segmentation values? Thanks for any advice on this issue.
why don't you try doing a within-group (ideally control) across-sequence study and see what the effect is. On Wed, 7 Oct 2009, James Root wrote:
In regard to group differences, roughly 1/3 of the scans in one group match the voxel size and orientation parameters in the other. The remaining scans are uniformly different from each other. Not knowing the specific effects that comparing unequal parameters will have in a Freesurfer analysis, is there a threshold or rule of thumb for when not to compare groups with different parameters? Are there specific parameter differences to avoid (x, y, z, size; orientation)? Thanks again for your help.
On Tue, Oct 6, 2009 at 12:14 PM, Bruce Fischl fischl@nmr.mgh.harvard.eduwrote:
Hi James,
it certainly can. If one group had all one scan type and another all a different one then you wouldn't be able to tell if the effects you find are due to the group membership or the scan differences.
cheers, Bruce
On Tue, 6 Oct 2009, James Root wrote:
I am planning an analysis using freesurfer to compare 2 groups. The SPGRs
that will be used vary in regard to slice thickness (from 0.86 to 1.5 in each axis), number of slices acquired (100-256), and orientation (axial; coronal) due to the fact that these were collected over a number of years on the same scanner. In another posting I read that this might be a problem but wanted to revisit this to see if any specific heterogeneity in the parameters should be particularly avoided. Also, if the groups differ such that one group has largely similar scan parameters and the second group is more heterogeneous, will this introduce any systematic bias in cortical thickness estimates, parcellation, or segmentation values? Thanks for any advice on this issue.
I will. Thanks Bruce.
On Oct 7, 2009 9:07am, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote:
why don't you try doing a within-group (ideally control) across-sequence study and see what the effect is.
On Wed, 7 Oct 2009, James Root wrote:
In regard to group differences, roughly 1/3 of the scans in one group match
the voxel size and orientation parameters in the other. The remaining scans
are uniformly different from each other. Not knowing the specific effects
that comparing unequal parameters will have in a Freesurfer analysis, is
there a threshold or rule of thumb for when not to compare groups with
different parameters? Are there specific parameter differences to avoid (x,
y, z, size; orientation)? Thanks again for your help.
On Tue, Oct 6, 2009 at 12:14 PM, Bruce Fischl fischl@nmr.mgh.harvard.edu>wrote:
Hi James,
it certainly can. If one group had all one scan type and another all a
different one then you wouldn't be able to tell if the effects you find are
due to the group membership or the scan differences.
cheers,
Bruce
On Tue, 6 Oct 2009, James Root wrote:
I am planning an analysis using freesurfer to compare 2 groups. The SPGRs
that will be used vary in regard to slice thickness (from 0.86 to 1.5 in
each axis), number of slices acquired (100-256), and orientation (axial;
coronal) due to the fact that these were collected over a number of years
on
the same scanner. In another posting I read that this might be a problem
but wanted to revisit this to see if any specific heterogeneity in the
parameters should be particularly avoided. Also, if the groups differ
such
that one group has largely similar scan parameters and the second group is
more heterogeneous, will this introduce any systematic bias in cortical
thickness estimates, parcellation, or segmentation values? Thanks for
any
advice on this issue.
freesurfer@nmr.mgh.harvard.edu
-
Bruce Fischl -
Douglas N Greve -
James Root -
jcroot＠gmail.com