I have a question about if it is possible to use freesurfer programs by first doing bbregister on output from 3dvolreg and then using mri_cvs_register for coregistration of the freesurfer anatomical to a CVS35MNI152 template and then use mri_vol2vol to get the BOLD in this space. Then it was suggested to take this output and use it in 3dDeconvolve and then 3dANOVA2 etc. Is it possible to do this? What we really want is to improve the resolutions and get surface based results in order to improve our ability to estimate the most significant cluster in the anterior right insula using BR regressor coefficient data from 3dDeconvolve so we can use its location for TMS. What do you suggest?
Thanks for all your help ahead of time.
You should be able to use any software package that produces NIFTI datasets. The only problem I can see is you will blur your data with the multiple interpolations at each transformation. Within AFNI, you can combine transformations, and I imagine FreeSurfer has some way to do that too, but combining transformations across packages is difficult.
I was wondering if you have to convert nifti to AFNI format in order for the output which is the Bold in standard space obtained from the freesurfer vol2vol program which follows from using mri_cvs_register to register the volumetric to the standard brain cvs_avg35_inMNI152? I try to use the .nii file in the AFNI viewer and try to do 3dAutomask on it but it gets hung. Then I tried to convert it to AFNI format using 3dcopy but it also gets hung. What I really want is to put the freesurfer result into 3dDeconvolve. Please let me know what I might be doing wrong on the AFNI side of things. I really appreciate your help.
Can you report on the exact command and results that are giving you problems?
Specifically none of the commands worked when using the freesurfer Bold. It just maybe that my nifti file derived from freesurfer is wrong. However, does AFNI commands allow nifti files as input? or do they need to be transformed using 3dcopy to AFNI format? Even this was not possible in this case. Also do AFNI commands in general allow the use of a freesurfer MNI152 template defined as ‘cvs_avg35_inMNI152’ or any other template other than AFNI templates such as TLRC to be used for example in 3dDeconvolve as input? I would like to use 3dDeconvolve with the freesurfer template if possible.
Please let me know I am very curious…
AFNI does read NIFTI datasets as input for virtually every command. Not sure what else to tell you other than to ask for the exact command and output.
Thanks! The AFNI commands were hung–they were just killed after a long while so I have no output. I am going to check the stuff from freesurfer .
I appreciate your help and I may get back to you in the future. -Linda
Thank You! for your attention. It turned out that the resulting Bold matrix from freesurfer is over 37GB and causes AFNI to be hung. Besides getting a better computer whose standards you can suggest, is it possible in anyway using AFNI to reduce the size by using runs instead ? However, I feel we would have a big problem with the deconvolution since we use the old version with impulse response functions and we use uniquely defined behavioral response regressors which were defined over all runs. Is it even possible to save ‘space’ since the runs should be concatenated in time in order to use 3dDeconvolve–isn’t that true? Certainly you can’t run a deconvolution on each run separately and combine them to get the same result as if they were concatenated? I am not a math expert but that seems like the case to me.
I appreciate your wisdom,
Our typical processing pipeline recommendations usually do not concatenate runs together. The data is still all analyzed with a single linear model that has the advantage of separate baseline estimations for each run. See afni_proc.py for examples and uber_subject.py for a simple graphical interface.
You say a single linear model, does that mean the runs are concatenated together to create this single model? Otherwise if you have three separate runs, for example, you don’t do a separate linear model for each doing 3 regression analyses (i.e. LS models) and then combine beta coefficients together from each run’s linear model somehow. I know baselines can be unique to each run. But you do concatenate runs together to do the deconvolution linear model or not? And if you don’t can you explain better?