I guess you could apply your transform to both the healthy controls and to your patient, can then create an FSGD file with two classes as before but no continuous variable. Then just have a contrast that computes the diff between the patient and the mean of the controls.
On 5/8/2020 11:49 AM, Xiaojiang Yang wrote:
External Email - Use Caution
Hi Douglas,
Could you please reply my last email (sent to freesurfer@nmr.mgh.harvard.edu mailto:freesurfer@nmr.mgh.harvard.edu) regarding my question about the weighted age? I am eager to listen to your answer/opinion to my question. For your convenience, I also include the whole email chain below this email.
Thank you very much!
Sincerely,
Xiao
https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=from :%22Douglas+N.+Greve%22 Douglas N. Greve https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=date :20200507 Thu, 07 May 2020 14:56:08 -0700 sorry, please include the previous email chain so that I know what you are referring to On 5/7/2020 5:33 PM, Xiaojiang Yang wrote: External Email - Use Caution Douglas, Just as I mentioned in the initial email, subjects in the healthy group have different ages. Because average cortical thickness is believed to be declined as the age gets older, I want to "normalize" the group by multiplying each subject's thickness with an age-dependent scalar (pre-calculated constant). Suppose I take age 30 as the "standard" age, then subjects with ages greater than 30 will have scalars greater than 1 (depend on age); subjects with ages smaller than 30 will have scalars less than 1. In this way, all subjects in the group looks like to be at the same age (30). Of course, I also suspect that there is no need to do the above mentioned normalization if I use mri_glmfit. But my initial intention was using non-linear scalars, and mri_glmft can only use linear fit. Please give me your opinion. Thank you! Xiao https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=from :%22Douglas+N.+Greve%22 Douglas N. Greve https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=date :20200507 Thu, 07 May 2020 14:06:27 -0700 I don't know what weight is. Can you elaborate? On 5/7/2020 4:54 PM, Xiaojiang Yang wrote: External Email - Use Caution Hi Douglas, Thank you very much! Could you please help me with more detailed information? 1)Should I use "Variables age" or "Variables age weight" in the fsgd file? If I can use either of them, which one is better? I don't know what weight is. Can you elaborate? 2)For the contrast file, the content "1 -1 0" you mentioned, it corresponds to the "Variables age" fsgd file, correct? Yes 3)If I use the "Variables age weight" fsgd file, what will be the content of the contrast file? Is it "1 -1 0 0"? Yes Thank you again! Xiao https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=from :%22Douglas+N.+Greve%22 Douglas N. Greve https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=date :20200507 Thu, 07 May 2020 11:55:27 -0700 I think it will work properly if you use the raw data as input and create an FSGD file with two classes: (1) the individual subject and (2) healthy subjects. Also include age as a covariate. Use DOSS instead of DODS. Then set your contrast to be 1 -1 0 and run mri_glmfit. I think that should properly account for age. On 5/7/2020 2:44 PM, Xiaojiang Yang wrote: External Email - Use Caution Dear FS experts, I have a group of healthy subjects. Given a new individual subject, I want to compare its cortical thickness to the healthy group so that I can find where the thickness is abnormal (thickening or thinning). I use fsaverage as the template subject to calculate mean and std of the healthy group. Since subjects in the healthy group have different ages, I do a "normalization" process to all the subjects in the group BEFORE calculating the mean and std. Thus the normalized healthy group can be regarded as all subjects having the same "standard" age. The normalization process is just multiplying some pre-obtained scale (varied by age, but not linearly ) to the file *?h.thickness.fwhm0.fsaverage.mgh* of each subject that are already calculated by "recon-all --qcache" command. So the mean and std of the group are actually weighted in the sense of Freesurfer's recon-all results. My questions are: 1)If I do vertex-wise t-tests by simply comparing individual subject's thickness to the group using mris_calc but WITHOUT using mri_glmfit, can I still use mri_glmfit-sim to do multiple comparisons correction? If yes, how? 2)If the answer is no for the above question, how should I use mris_glmft to implement my above thoughts so that I can use use mri_glmfit-sim? Specifically, how can I use the weighted mean for the group? 3)Take a step back, if I consider the weight is linearly correlated with the age, how to design the fsdg and contrast file? Thank you very much! Xiao
External Email - Use Caution
Thank you Douglas! If I apply my transform to both the healthy controls and the patient, I guess I would lose the convenience of using mri_glmfit, which eventually loses the convenience of using mri-glmfit-sim (I don't care not using mri_glmfit, but I want to use mri_glmfit-sim to do multiple comparisons correction). That is because mri_glmfit uses the prepared output from mris_preproc, which I have no control over to use the transformed subject. The only way I can circumvent this is to use my transformed thickness file to overwrite the FS calculated thickness file, but that is what I am hesitant to do. If you have good suggestions regarding this, please let me know. Thanks! Xiao
On Mon, May 11, 2020 at 12:25 PM Douglas N. Greve dgreve@mgh.harvard.edu wrote:
I guess you could apply your transform to both the healthy controls and to your patient, can then create an FSGD file with two classes as before but no continuous variable. Then just have a contrast that computes the diff between the patient and the mean of the controls.
On 5/8/2020 11:49 AM, Xiaojiang Yang wrote:
External Email - Use CautionHi Douglas,
Could you please reply my last email (sent to freesurfer@nmr.mgh.harvard.edu) regarding my question about the weighted age? I am eager to listen to your answer/opinion to my question. For your convenience, I also include the whole email chain below this email.
Thank you very much!
Sincerely,
Xiao
<https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=f...
:%22Douglas+N.+Greve%22> Douglas N. Greve
<https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=d...
:20200507> Thu, 07 May 2020 14:56:08 -0700
sorry, please include the previous email chain so that I know what you are
referring to
On 5/7/2020 5:33 PM, Xiaojiang Yang wrote:
External Email - Use CautionDouglas,
Just as I mentioned in the initial email, subjects in the healthy group have
different ages. Because average cortical thickness is believed to be
declined as the age gets older, I want to "normalize" the group by
multiplying each subject's thickness with an age-dependent scalar
(pre-calculated constant). Suppose I take age 30 as the "standard" age, then
subjects with ages greater than 30 will have scalars greater than 1 (depend
on age); subjects with ages smaller than 30 will have scalars less than 1.
In this way, all subjects in the group looks like to be at the same age
(30).
Of course, I also suspect that there is no need to do the above mentioned
normalization if I use mri_glmfit. But my initial intention was using
non-linear scalars, and mri_glmft can only use linear fit. Please give me
your opinion. Thank you!
Xiao
<https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=f...
:%22Douglas+N.+Greve%22> Douglas N. Greve
<https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=d...
:20200507> Thu, 07 May 2020 14:06:27 -0700
I don't know what weight is. Can you elaborate?
On 5/7/2020 4:54 PM, Xiaojiang Yang wrote:
External Email - Use CautionHi Douglas,
Thank you very much! Could you please help me with more detailed
information?
1)Should I use "Variables age" or "Variables age weight" in the fsgd file?
If I can use either of them, which one is better?
I don't know what weight is. Can you elaborate?
2)For the contrast file, the content "1 -1 0" you mentioned, it corresponds
to the "Variables age" fsgd file, correct?
Yes
3)If I use the "Variables age weight" fsgd file, what will be the content of
the contrast file? Is it "1 -1 0 0"?
Yes
Thank you again!
Xiao
<https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=f...
:%22Douglas+N.+Greve%22> Douglas N. Greve
<https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=d...
:20200507> Thu, 07 May 2020 11:55:27 -0700
I think it will work properly if you use the raw data as input and create an
FSGD file with two classes: (1) the individual subject and (2) healthy
subjects. Also include age as a covariate. Use DOSS instead of DODS. Then
set your contrast to be 1 -1 0 and run mri_glmfit. I think that should
properly account for age.
On 5/7/2020 2:44 PM, Xiaojiang Yang wrote:
External Email - Use CautionDear FS experts,
I have a group of healthy subjects. Given a new individual subject, I want
to compare its cortical thickness to the healthy group so that I can find
where the thickness is abnormal (thickening or thinning).
I use fsaverage as the template subject to calculate mean and std of the
healthy group. Since subjects in the healthy group have different ages, I do
a "normalization" process to all the subjects in the group BEFORE
calculating the mean and std. Thus the normalized healthy group can be
regarded as all subjects having the same "standard" age. The normalization
process is just multiplying some pre-obtained scale (varied by age, but not
linearly ) to the file *?h.thickness.fwhm0.fsaverage.mgh* of each subject
that are already calculated by "recon-all --qcache" command. So the mean and
std of the group are actually weighted in the sense of Freesurfer's
recon-all results.
My questions are:
1)If I do vertex-wise t-tests by simply comparing individual subject's
thickness to the group using mris_calc but WITHOUT using mri_glmfit, can I
still use mri_glmfit-sim to do multiple comparisons correction? If yes, how?
2)If the answer is no for the above question, how should I use mris_glmft to
implement my above thoughts so that I can use use mri_glmfit-sim?
Specifically, how can I use the weighted mean for the group?
3)Take a step back, if I consider the weight is linearly correlated with the
age, how to design the fsdg and contrast file?
Thank you very much!
Xiao
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 Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
Just call your thickness file some other name, eg, lh.xformed.thickness.mgz, the just specify --meas xformed.thickness.mgz
On 5/11/2020 3:52 PM, Xiaojiang Yang wrote:
External Email - Use Caution
Thank you Douglas! If I apply my transform to both the healthy controls and the patient, I guess I would lose the convenience of using mri_glmfit, which eventually loses the convenience of using mri-glmfit-sim (I don't care not using mri_glmfit, but I want to use mri_glmfit-sim to do multiple comparisons correction). That is because mri_glmfit uses the prepared output from mris_preproc, which I have no control over to use the transformed subject. The only way I can circumvent this is to use my transformed thickness file to overwrite the FS calculated thickness file, but that is what I am hesitant to do. If you have good suggestions regarding this, please let me know. Thanks! Xiao
On Mon, May 11, 2020 at 12:25 PM Douglas N. Greve <dgreve@mgh.harvard.edu mailto:dgreve@mgh.harvard.edu> wrote:
I guess you could apply your transform to both the healthy controls and to your patient, can then create an FSGD file with two classes as before but no continuous variable. Then just have a contrast that computes the diff between the patient and the mean of the controls. On 5/8/2020 11:49 AM, Xiaojiang Yang wrote:External Email - Use Caution Hi Douglas, Could you please reply my last email (sent to freesurfer@nmr.mgh.harvard.edu <mailto:freesurfer@nmr.mgh.harvard.edu>) regarding my question about the weighted age? I am eager to listen to your answer/opinion to my question. For your convenience, I also include the whole email chain below this email. Thank you very much! Sincerely, Xiao <https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=from :%22Douglas+N.+Greve%22> Douglas N. Greve <https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=date :20200507> Thu, 07 May 2020 14:56:08 -0700 sorry, please include the previous email chain so that I know what you are referring to On 5/7/2020 5:33 PM, Xiaojiang Yang wrote: External Email - Use Caution Douglas, Just as I mentioned in the initial email, subjects in the healthy group have different ages. Because average cortical thickness is believed to be declined as the age gets older, I want to "normalize" the group by multiplying each subject's thickness with an age-dependent scalar (pre-calculated constant). Suppose I take age 30 as the "standard" age, then subjects with ages greater than 30 will have scalars greater than 1 (depend on age); subjects with ages smaller than 30 will have scalars less than 1. In this way, all subjects in the group looks like to be at the same age (30). Of course, I also suspect that there is no need to do the above mentioned normalization if I use mri_glmfit. But my initial intention was using non-linear scalars, and mri_glmft can only use linear fit. Please give me your opinion. Thank you! Xiao <https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=from :%22Douglas+N.+Greve%22> Douglas N. Greve <https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=date :20200507> Thu, 07 May 2020 14:06:27 -0700 I don't know what weight is. Can you elaborate? On 5/7/2020 4:54 PM, Xiaojiang Yang wrote: External Email - Use Caution Hi Douglas, Thank you very much! Could you please help me with more detailed information? 1)Should I use "Variables age" or "Variables age weight" in the fsgd file? If I can use either of them, which one is better? I don't know what weight is. Can you elaborate? 2)For the contrast file, the content "1 -1 0" you mentioned, it corresponds to the "Variables age" fsgd file, correct? Yes 3)If I use the "Variables age weight" fsgd file, what will be the content of the contrast file? Is it "1 -1 0 0"? Yes Thank you again! Xiao <https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=from :%22Douglas+N.+Greve%22> Douglas N. Greve <https://www.mail-archive.com/search?l=freesurfer@nmr.mgh.harvard.edu&q=date :20200507> Thu, 07 May 2020 11:55:27 -0700 I think it will work properly if you use the raw data as input and create an FSGD file with two classes: (1) the individual subject and (2) healthy subjects. Also include age as a covariate. Use DOSS instead of DODS. Then set your contrast to be 1 -1 0 and run mri_glmfit. I think that should properly account for age. On 5/7/2020 2:44 PM, Xiaojiang Yang wrote: External Email - Use Caution Dear FS experts, I have a group of healthy subjects. Given a new individual subject, I want to compare its cortical thickness to the healthy group so that I can find where the thickness is abnormal (thickening or thinning). I use fsaverage as the template subject to calculate mean and std of the healthy group. Since subjects in the healthy group have different ages, I do a "normalization" process to all the subjects in the group BEFORE calculating the mean and std. Thus the normalized healthy group can be regarded as all subjects having the same "standard" age. The normalization process is just multiplying some pre-obtained scale (varied by age, but not linearly ) to the file *?h.thickness.fwhm0.fsaverage.mgh* of each subject that are already calculated by "recon-all --qcache" command. So the mean and std of the group are actually weighted in the sense of Freesurfer's recon-all results. My questions are: 1)If I do vertex-wise t-tests by simply comparing individual subject's thickness to the group using mris_calc but WITHOUT using mri_glmfit, can I still use mri_glmfit-sim to do multiple comparisons correction? If yes, how? 2)If the answer is no for the above question, how should I use mris_glmft to implement my above thoughts so that I can use use mri_glmfit-sim? Specifically, how can I use the weighted mean for the group? 3)Take a step back, if I consider the weight is linearly correlated with the age, how to design the fsdg and contrast file? Thank you very much! XiaoThe 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 Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Douglas N. Greve -
Xiaojiang Yang