Working with realigned & unwarped images in DPARSFA

Submitted by Mamtis13 on Wed, 09/04/2013 - 20:06

Dear members,

I am relatively new to resting state analysis and the DPARSFA toolbox, any comments, suggestions and feedback to the following questions would be highly appreciated!

Has someone worked with fieldmaps in SPM, realigned & unwarped in SPM and then continued the preprocessing and FC analysis in DPARSFA? If so, what are the options required to be checked in DPARSFA to continue the preprocessing? I'd like to get some advice on how you organised your workflow (i.e. your pre-processing steps).

I have resting state data (2 runs/person) and field map information. I discovered that there were substantial field inhomogeneities which I needed to correct for using subjects' field maps. DPARSFA cannot (at present) incorporate this step into the preprocessing, which is why I have used the Fieldmap toolbox in SPM8 to generate vdm files, and have used these to realign & unwarp my data. (It seems not possible to only unwarp in SPM, hence I also had to realign outside of DPARSFA.) I would now like to use the realigned, unwarped images for the rest of the preprocessing and functional connectivity analysis in DPARSFA, but am unsure which options to check for preprocessing given that my images are already realigned & unwarped. 

Can I leave out the realign option in DPARSFA and still obtain valid results after analysis? It appears that, when I leave realign unchecked but check all other options, the code crashes during the next step. During the last step of preprocessing, covariates are regressed out. With realign being left unchecked, no motion parameters are being regressed out, only CSF and WM (if ticked). Is it true that, because of the unwarping step, motion parameters do not need to be regressed out during the last step of preprocessing?

Apologies for the long query, as I said I am new to this and am still learning. Any advice would be greatly appreciated!

Cheers, Mamtis

Submitted by ZHANG_RESTadmin on Thu, 09/05/2013 - 11:40

 during DARSF doing realignment, it doesn't only generate new ra*.img/hdr in FunImgAR directory, but also generates realign parameters (6 head motion curves) and mean functional images in RealignParameter directory with mean* and rp_* as prefix. You should create the same dir structure and store all necessary files in to make the rest procedure works fine.   To do this, you can firstly do a pilot processing that does not include filedmap (i.e., ordinary pipeline), then you will know how the file and dir organized.  

Submitted by Mamtis13 on Sun, 09/08/2013 - 17:01

Hello,

Thank you very much for the response! I tried this before but unfortunately even when I recreate the directory structure and store the realignment parameter files in the RealignmentParameters folder, the procedure crashes on the last step (regressing out covariates), which I think is the result of there not being a voxel-specific head motion procedure? Does the toolbox generate voxel-specific head motion parameters as part of the realign procedure (i.e. are they intrinsically linked) or can voxel-specific head motion parameters be estimated based on rp_* and mean* files only?

On a more theoretical level, can you tell me if motion parameters should not be regressed out during the last step of preprocessing if I am working with unwarped images?

Thank you very much in advance for your help!
Mamtis

Submitted by ZHANG_RESTadmin on Thu, 09/12/2013 - 10:42

 you mean you need to include voxel specific headmotion in the nuisance regressors? if so, you need to calculate them out first.
they are not rp_txt or mean file though.   

2nd question, in my opinion, unwarp did some correction based on the fieldmap but it never correct the head motion and the effect caused by it. so you still have to do the head motion regression. but, anyway, usually they do it only using 6 rigid transformation parameters wihch is your rp file, and voxel specific head motion can not be calculated from it.