Forum

Notifications
Clear all

Correction Motion Artifacts - Day4 of 2020 course  


David Boas
Posts: 21
(@dboas)
Eminent Member
Joined: 2 months ago
 

Lisa raised several general questions about correcting MA as discussed in Day 4 of the fnirs course 2020.

  1. Is it inappropriate or problematic to run multiple types (i.e., 2+) of motion correction on a single run/subject? I assume too many corrections will result in lost brain activation signal, but sometimes it seems a single correction is not reducing enough MA.
  2. Is it okay to run different types of motion correction across runs/subjects? (Subject A gets better HRF with tPCA, but Subject B with Spline, for example). Is it okay to do different MA corrections prior to running the group averages or should the previous steps be universal across files (runs or subjects)?
  3. How is the above (#2) accomplished? By selecting the run/subject in ‘Current Processing Element’, editing the parameters on the processing stream, then running at that level? Does ‘apply to all’ need to be unchecked? After individually processing the separate files, will running at the Group level with ‘apply to all’ checked apply the most recent run processing stream to all the subjects and thereby negate any individualization that was done?
2 Replies
David Boas
Posts: 21
(@dboas)
Eminent Member
Joined: 2 months ago

Re question 1:
We strongly encourage you to use only one motion correction method when analyzing your data. You can try different methods, one at a time, to visualize the impact it has on your data. But at the end, choose only one to use.

Re question 2:
You really should use the same processing stream and function parameters for all subjects. Tailoring your processing stream functions and parameters to get the "optimal" result for each run means that you are introducing your own subjective bias into the analysis to get the result you want.

Re question 3:
This is why we have "apply all" checked, to make sure you use the same processing parameters for all subjects. Of course, you can uncheck this and then "optimize" parameters for each subject. But then you are at risk for producing the results you want to see. This can easily bias the results of your study to achieve your expected result.

Reply
1 Reply
Lisa Bunker
(@lbunk)
Joined: 1 month ago

New Member
Posts: 1

@dboas Thank you!

Reply
Share: