13 Oct
2015
13 Oct
'15
9:47 a.m.
Hi Matthieu,
1. I think it is OK, maybe Jorge can comment, if not. 2. Not really. You could do a linear fit within each subject separately (at each location) and then do a Wilcoxon Ranksum (or a Signed Rank ) Test. The non-parametric testing will help with potential outliers, the problem still is that you have differently many time points per subjects, and that is not considered in that analysis at all. I am not a statistician, so really the wrong person to talk to about that.
Best, Martin
On 10/13/2015 09:37 AM, Matthieu Vanhoutte wrote:
Hello Martin,
Many thanks for your answer !
Indeed, I didn't add age and gender as covariates according the fact of my small number of patients group. Moreover, I couldn't put the slope as a random effect because of no convergence problem...
Concerning the convergence, at the end of the estimation model with only one random effect on intercept, I got this error :
/Summary: Algorithm did not converge at 9.1964 percent of the total number of locations. Total elapsed time is 46.692 minutes./
- Is this challenging me about the validity of the results ?
- Do you know a mean to avoid or take account of spurious effects
(due to violation of the assumption of gaussian noise) in the statistical analysis results ?
Best regards,
Matthieu
2015-10-13 15:26 GMT+02:00 Martin Reuter <mreuter@nmr.mgh.harvard.edu mailto:mreuter@nmr.mgh.harvard.edu>:
Hi Matthieu, 1. The spatiotemporal approach is designed for surfaces, not for voxel images, so regular mass univariate (vw). 2. Yes, that can be done with LME. 1 group is a simple design: column of 1, colum of time_from_baseline, other covariates such as age (I would not include too many as you only have very few subjects). Then set the intercept as random effect and test for a slope effect (2nd column). You can also play around with setting slope as random effect, but with only 7 subjects, not sure. You can do a model comparison as described on our wiki. 3. Not sure if I would include gender (unless there is good reason for that). You are looking at within subject slopes. I don't think you need to demean age. 4. does not matter. If you look at the betas they will be in unit/days or unit/years (or week , month) depending on your decision. The danger with such a small group is that outliers will affect your results a lot (they severely violate the assumption of gaussian noise) and can produce spurious effects. Best, Martin On 10/11/2015 04:31 PM, Matthieu Vanhoutte wrote:Dear experts, Could anyone please help me ? My group has small size (7 subjects) with non equally spaced different timepoints (from 2 up to 4). Best regards, Matthieu Le 30 sept. 2015 13:57, "Matthieu Vanhoutte" <matthieuvanhoutte@gmail.com <mailto:matthieuvanhoutte@gmail.com>> a écrit : Hi Martin, Thank you for helping ! 1) What should I use for the parameter estimation: "mass univariate" or Novel mass-univariate tools (spatiotemporal models)" ? 2) I have one group of subjects with different timepoints (non equally spaced and different numbers of timepoints), and would like to see the significant change of FA according to time. Will LME study allow me to get this change and how to define contrast to assess it ? Are the results of this longitudinal study directionals, i.e. is it possible to establish an increase or decrease of FA in significant regions according to time ? 3) Should the covariates (age, gender...) be demeaned or centered ? 4) Has the accurate time variable to be in years absolutely or could it be also in days ? Best regards, Matthieu 2015-09-29 17:11 GMT+02:00 Martin Reuter <mreuter@nmr.mgh.harvard.edu <mailto:mreuter@nmr.mgh.harvard.edu>>: Hi Matthieu, use lme_mass_fit_vw and the y is a simple vector. I would mask the image before doing this and only pass the brain voxels (to increase speed and reduce comparisons). Best, Martin On 09/29/2015 10:43 AM, Matthieu Vanhoutte wrote:Hi Martin, I understand for the multiple comparisons problem. But if I consider a voxel-by-voxel analysis just as hippocampal volume (lme_fit_FS for estimation), how could I consider a correction for multiple comparisons in this univariate case ? Shouldn't I use the lme_mass_fit or lme_mass_fit_vw functions in my mass univariate case ? If yes, how do I have to format my 3D FA volume as input of the LME toolbox ? Best regards, Matthieu ------------------------------------- Matthieu Vanhoutte, MSc Research Engineer - Department of Neuroradiology Regional University Hospital, Lille, France 2015-09-29 16:34 GMT+02:00 Martin Reuter <mreuter@nmr.mgh.harvard.edu <mailto:mreuter@nmr.mgh.harvard.edu>>: Hi Matthieu, yes, multiple comparisons are a problem for any mass univariate approach. You can use the FDR2 correction (in the lme matlab tools) which is less conservative than standard FDR. You can also work in specific ROI's and average there, to reduce the number of comparisons. Best, Martin On 09/29/2015 09:11 AM, Matthieu Vanhoutte wrote:Hi Martin, Thank you for your answer. However, I would like to consider FA maps as a mass-univariate problem and not a univariate voxel-by-voxel as hippocampal volume. Indeed, my voxels aren't independant, are they ? So, how to correct for multiple comparisons then ? Best regards, Matthieu 2015-09-29 15:04 GMT+02:00 Martin Reuter <mreuter@nmr.mgh.harvard.edu <mailto:mreuter@nmr.mgh.harvard.edu>>: Hi Matthieu, if all your images are perfectly registered, you can do LME on a voxel-by-voxel basis, just as if you had hippocampal volume or any other ROI measure. Best, Martin On 09/29/2015 06:46 AM, Matthieu Vanhoutte wrote:Dear experts, I would like to know if it is possible to make a longitudinal study with LME toolbox from volume FA maps registered in a common space ? I don't have T1 images so the recon-all process couldn't be processed. But if I put my 3D FA volume of dim = [nx,ny,nz] in a 1D nx*ny*nz voxels format, could it be in good way for LME ? Thanks for helping ! Best regards, Matthieu _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer-- Martin Reuter, PhD Assistant Professor of Radiology, Harvard Medical School Assistant Professor of Neurology, Harvard Medical School A.A.Martinos Center for Biomedical Imaging Massachusetts General Hospital Research Affiliate, CSAIL, MIT Phone:+1-617-724-5652 <tel:%2B1-617-724-5652> Web :http://reuter.mit.edu _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu <mailto: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 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 <mailto:Freesurfer@nmr.mgh.harvard.edu> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer-- Martin Reuter, PhD Assistant Professor of Radiology, Harvard Medical School Assistant Professor of Neurology, Harvard Medical School A.A.Martinos Center for Biomedical Imaging Massachusetts General Hospital Research Affiliate, CSAIL, MIT Phone: +1-617-724-5652 <tel:%2B1-617-724-5652> Web : http://reuter.mit.edu _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu <mailto: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 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 <mailto:Freesurfer@nmr.mgh.harvard.edu> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer-- Martin Reuter, PhD Assistant Professor of Radiology, Harvard Medical School Assistant Professor of Neurology, Harvard Medical School A.A.Martinos Center for Biomedical Imaging Massachusetts General Hospital Research Affiliate, CSAIL, MIT Phone: +1-617-724-5652 <tel:%2B1-617-724-5652> Web : http://reuter.mit.edu _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu <mailto: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 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 <mailto:Freesurfer@nmr.mgh.harvard.edu> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer-- Martin Reuter, PhD Assistant Professor of Radiology, Harvard Medical School Assistant Professor of Neurology, Harvard Medical School A.A.Martinos Center for Biomedical Imaging Massachusetts General Hospital Research Affiliate, CSAIL, MIT Phone:+1-617-724-5652 <tel:%2B1-617-724-5652> Web :http://reuter.mit.edu _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu <mailto: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 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
--
Martin Reuter, PhD
Assistant Professor of Radiology, Harvard Medical School
Assistant Professor of Neurology, Harvard Medical School
A.A.Martinos Center for Biomedical Imaging
Massachusetts General Hospital
Research Affiliate, CSAIL, MIT
Phone: +1-617-724-5652
Web : http://reuter.mit.edu