About preprocessing the resting state fMRI data

Submitted by YukiSakai on Sat, 07/18/2009 - 00:01

Hi, REST exparts
 
 Thank you for releasing convenient tool called DPARSF. I have some questions about that.

1. When I have tried to run DPARSF on my Mac. PC, I always met the error said ' time points of your data are 0'. I putted 195 volumes EPI Nifti data of each subjects into one folder and then putted these folders of each subjects into one folder named 'FunImgNormalizedSmoothed'. If I tried to run DPARSF on my Win. PC, same data was recognized as 195 volumes, but I couldn't run because the computer load was out of memory. Is there the way to run DPARSF on Mac. PC?

2. I understand that I have to set covariates on each subjects separately in order to run FC analysis on REST tool. If I run DPARSF, I guess I'll get regressed out hdr. img. data of each subjects. Do this process mean that each data were regressed out by their own covariates? Do this mean each voxel have detrended, filtered and regressed out time course data?  If so, is it possible that we can run FC analysis at one time using this regressed out data?

 And I have one question about slice viewer of REST tool. Now I preprocess my EPI data with FSL and SPM5 as below.
   stripping of non-brain structures
 　realign(register to first)
   slice timing correction
   normalize to EPI template (voxel size 2x2x2) 
   smoothing (FWHM 6x6x6)
Then, I run FC analisis on REST tool and get zFCmap.
But when I see this zFCmap on slice viewer of REST tool, sagittal view was not match to axial view. I mean some part of anterior-inferior region of image (e.g. OFC) exist only in sagittal view but not in axial view. When I see same image with SPM5, this kind of mismatch didn't happen. Why do this kind of mismatch happen?

Thank you in advance for your help. Any suggestion will be really appreciate.
Sincerely,
 Yuki

Submitted by YAN Chao-Gan on Sat, 07/18/2009 - 23:30

Hi!
1. Have you set the "Time Points" parameter in the up-right panel in DPARSF? It is just at the top of the parameter "TR". If you encountered the error of "time points of your data are 0' even you set the Time Points parameter, could you help me to debug it? I do not have a Mac system, I just tested DPARSF in Windows and Linux. Also, if you want to resolve the problem of "out of memory" in Windows, you may need to turn on the 3GB switch of Windows XP, see more details in http://www.microsoft.com/whdc/system/platform/server/PAE/PAEmem.mspx.
2. Yes, your guesses are right. You can run FC analysis at one time using this regressed out data. I strongly recommend you download a multimedia course from http://restfmri.net/forum/Course, this will keep you out of many unnecessary errors.
3. Thank you for your report. Ms An Li also report the error of sagittal view (the axial view is correct) a couple of days before. We have revised this problem, but we can not release it publicly now since we still working on revising and adding some functions to REST. If you are interested in it, we can send a closed test version to your email box.

Best wishes!

Submitted by YukiSakai on Thu, 07/30/2009 - 12:34

Hi,
 I'm sorry I'm really late.

1. I set "Time Points" parameter correctly, but error message always appears on my Mac. PC. By the way, I deeply appreciate your information about 3GB switch. I could run DPARSF on my Win. PC.
2. A multimedia course was really helpful to me.
3. It is good news to me. Would you mind sending the closed version to my email box?

There are another question about zFCmap.
4. If I have zFCmaps from one seed ROI of mant subjects, which way is the best to show distribution pattern of all subjects? Now I do one-sample t-test with SPM5 and get the distribution pattern. But, in this way, I can get only positive correlateion map but negative one. Which way is the best in order to make positeve and negative correlation map?

Thank you in advance for your help. Any suggestion will be really appreciate.
Sincerely,
 Yuki

Submitted by YAN Chao-Gan on Sat, 08/01/2009 - 11:08

Hi!
The closed test version of REST has been sent to your email box, please check it.

For your question, I think one-sample t-test  is a suitable way. I think you did not get the negative connectivity is caused by your missing of removing the global trend. It is reported that negative connectivity is revealed only after removal of global trend (Murphy et al., 2009). Please refer to Slice 63 to 75 in the course to get information for how to remove the global trend: http://www.restfmri.net/forum/Course

Murphy K, Birn RM, Handwerker DA, Jones TB, Bandettini PA (2009) The impact of global signal regression on resting state correlations: are anti-correlated networks introduced? Neuroimage 44:893-905.

Submitted by YukiSakai on Sun, 08/02/2009 - 02:40

Hi,
 Thank you very much for sending me The closed test version of REST. I'll try it soon.

 I know the fact about global signal regression and negative connectivity. Now, I preprocess my data with SPM5(realign, slice timing correction, normalize, smooth),  detrend and band pass filter with REST, regressed out 9 nuisance covariates from detrended and filtered data with DPARSF, and then calculate functional connectivity with REST. I use Covremoved file at the last step, so I assume it is no probem. I think blue color in zFCmap mean negative connectivitiy, is it alright?
 I mean I can't distinguish negative connectivoty from positive one, because those are same red gradation color in the outcome of one-sample t-test.
 And also when I do two-sample t-test, I can get significant region between healthy subjects and patients. But, in which way can I distinguish positive correlation region from negative one in the outcome region?
 I'm sorry it may be very basic question.

Thank you in advance for your help. Any suggestion will be really appreciate.
Sincerely,
 Yuki

Submitted by YAN Chao-Gan on Tue, 08/04/2009 - 17:12

You can find a text field in the up left panel of rest_sliceviewer, it will display the value of the current voxel, thus you can see the T value of that point.
You can also view your results by using MRIcroN, xjview or SPM5.
I planed to release the Part 2 of the Multimedia Course in this September, mainly about:
1. How to perform the statistical analysis based on FC, ReHo, ALFF and fALFF results.
2. How to view your results and make beautiful pictures.
I hope I can finish it then, thank you for your patience.

Submitted by YukiSakai on Mon, 08/17/2009 - 07:20

Hi,
 I want to ensure some questions.
1. About DPARSF trouble on Mac. PC, is it possible to use it now or is there the another way to use ?

2.Now, I preprocess my data with SPM5(realign, slice timing correction, normalize, smooth),  detrend and band pass filter with REST, regressed out 9 nuisance covariates from detrended and filtered data with DPARSF, and then calculate functional connectivity with REST. I use Covremoved file at the last step, so I assume it is no probem. I think red and blue color in zFCmap mean positive and negative connectivitiy, respectively. Is this understanding correct?

3.I know we can see t-value on slceiviewer and I'm looking forward to seeing the Part 2 of the Multimedia Course. But, I want to know only about the way to distinguish positive and negative correlation in the outocome of statistic analysis as soon as possible. I'm really sorry to trouble you all.

Thank you in advance for your help. Any suggestion will be really appreciate.
Sincerely,
 Yuki

Submitted by YAN Chao-Gan on Thu, 08/27/2009 - 21:29

Hi！
Sorry for the later reply since I have gone out for a vacation.
1. I do not have any Mac PC, thus it’s difficult for me to test DPARSF on Mac OS. Could you give me more details for this error in Mac OS? Please paste the info in the command window of MATLAB here, thanks a lot! I will try to make DAPRSF suitable in Mac OS.
2. Yes! Red means positive and Blue means negative if you use the default settings. You also can see this information from the color bar which is on the right side.
3. I hope I can release the Part 2 of the Multimedia Course in this September, thank you for your patience.
Best wishes!