How best to handle subjects with different number of runs?

Submitted by eglee on Fri, 03/24/2017 - 14:44

Hello All,

I am using DPARSF to process my resting state data which will primarily be used in seed to whole brain analyses .

I have about 15 subjects that have been scanned before/after an intervention.

12 of those subjects have two resting state scans for each visit, but three of the subjects only have one scan for one of their visits.

What is the best way to handle this in DPABI?

What I have currently been doing is splitting all the data up so that each run is thought of as a different subject.

After DPARSF runs, I then concatenate the results for all scans that took place in the same session.

Is this a fine way of doing this, or would there be a better alternative? Maybe I should also mention that each run is 120 timepoints (after dropping the first 4) with a TR of 3.

 

Thanks!

Forums:

YAN Chao-Gan

Mon, 04/10/2017 - 10:52

Hi,

I think "splitting all the data up so that each run is thought of as a different subject" is not a good practice.

You can select a scan (e.g., the first) for each subject.

Or you can average the FC (or ALFF...) maps across the two sessions for each subject.

eglee

Thu, 04/20/2017 - 22:13

Thanks for your reply!

In our setup, there are only a few seconds between the two functional scans.

What would be the advantage of averageing the functional connectivity maps of two scans over concatenating the scans and then calculating the functional connectivity maps?

Also any thoughts on if it is better to just take one scan over averaging two if you don't have two for all subjects?

Thanks in advance. Also if anyone knows any publications that talk about this it would be much appreciated!

YAN Chao-Gan

Thu, 04/27/2017 - 06:18

There might be a baseline change from one scan to another scan, which could inflate the correlation.

You can choose one (e.g., the 1st) if not all subjects have 2 scans.