External Email - Use Caution
Dear Dr Greve and Matt, your valable responses and explanations are highly appreciated!! Thank you so much!
From: freesurfer-bounces@nmr.mgh.harvard.edu freesurfer-bounces@nmr.mgh.harvard.edu on behalf of Glasser, Matthew glasserm@wustl.edu Sent: Sunday, August 5, 2018 6:01 PM To: Freesurfer support list Subject: Re: [Freesurfer] SNR covariate in cortical thickness & volumetrics between groups analyses
If you have only two protocols “high quality” and “low quality", you could presumably include a protocol variable as a covariate of no interest, thereby removing any effects that correlate with protocol. For this to work ideally, you would have similar numbers of low and high quality protocols across your comparison groups. I agree with Doug that SNR will generally tend to make results more variable, but not necessarily in a specific direction.
Matt. From: freesurfer-bounces@nmr.mgh.harvard.edu on behalf of Douglas Greve dgreve@mgh.harvard.edu Reply-To: Freesurfer support list freesurfer@nmr.mgh.harvard.edu Date: Sunday, August 5, 2018 at 4:48 PM To: "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] SNR covariate in cortical thickness & volumetrics between groups analyses
I thought I answered this question last week. See my rsponse below
I don't think it is correct conceptually. By including SNR as a covariate, you are saying that you expect the thickness to increase with higher SNR and decrease with lower SNR. This does not make sense to me. It sounds like you are trying to do a mixed effects analysis where you weight by the SNR. Still, mathematically, this is not a true mixed effect analysis (which requires estimates of the within and between subject noise). `
On 8/3/18 10:57 AM, John Anderson wrote:
‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On August 5, 2018 1:04 PM, John Anderson John.anderso@protonmail.com wrote:
Dear FS experts, I would like to inquire about including SNR as a covariate for the between groups comparison in cortical thickness. Is this procedure valid? I have some T1 images of low quality and the majority are high in quality. I reviewed the segmentation and the cortical surfaces and fixed manually as possible. I don't want to exclude images from the data sets, so I am thinking of regressing out SNR. It seems that the command "mri_segstats" can report SNR for every ROI in the brain. Is it correct to include whole brain SNR covariate for between group comparisons (cortical thickness and volumetric analyses). If yes, do I need to include actual SNR values in the FSGD file or the demeaned values ?
We appreciate your guidance! John