Questions on DPARSFA

Submitted by Rita on Wed, 03/30/2011 - 02:16

1. Should I use smReHO - 1 or just smReHO for a paired t-test?
2. In ExcludeSubjects.txt what is the default or recommended criteria for excluding subjects?
3. Which of the DPARSFA choices does NOT effect ReHo analysis? For example, if I check "smooth" will it smooth the data before ReHo or does "smooth" only apply to fALFF, ALFF, and FC analysis?

Submitted by dongzy08 on Sat, 04/02/2011 - 09:17

1. Both are right, the results of a paired t-test are same no matter which you use.
2. There is not a accurate criterion. It depends on experimental conditions, you see the criteria for children and adults are different. But the 3 mm and 3 degree is the maximum for analysis.
3. It will do smooth before ReHo, if you check it.

Submitted by Rita on Tue, 04/05/2011 - 03:53

Thanks for the response!
2 more questions:
1. DPARSFA seems to work fine if I have .img files in FunImg. Is that fine to use?
2. Does "Regress out Nuissance Covariates" apply at all to ReHo? Or is it just for FC analysis?

Submitted by dongzy08 on Tue, 04/05/2011 - 15:21

1. Yes, DPARSFA could start at any step of data processing.
2. Just for FC analysis.

Submitted by Rita on Fri, 04/08/2011 - 01:25

Thanks again for the reply!
1 more question:
What is the recommended cluster size? I know in the paper introducing ReHo that was still undecided. Has there been any consensus on which to use? Is there a list of pros and cons to each?

Submitted by dongzy08 on Sun, 04/10/2011 - 20:50

They have not recommended the cluster size. In Zang's paper (Zang et al., 2004), they listed the number of masses showing significant difference between two different states. They find more masses and voxels by the cluster size =27. Further, Zang's group has discovered that there were more voxels showing significant difference between eye open (EO) resting-state and eye closed (EC) resting-state by the cluster size =27 (Dr. Liu's thesis, 2010). But there are not enough arguments to prove which cluster size is the best. If you are interested in it, you may have a try in different cluster size and compare the results just like Zang's methods (Zang et al., 2004),.