Hi Alfonso,

I have a question regarding the QA plots of denoising.

I have 164 subjects and the distribution of FC values looks fine (please see attached). However, I cannot get the QC-FC above 81% match with null hypothesis (for maxmotion, mean motion, and MeanGSchange).
I have already increased the white matter and CSF components to 16 and 10, respectively.

I have also a few subjects with very high scrubbing values. I believe that those subjects are impacting the QA plots.
Given that I have already increased the WM and CSF components, is there a way to further improve the QC-FC percentage or should I just remove my subjects with high scrubbing values?

Any guidance would be very appreciated.

Many thanks. 

Hugo

Hi Hugo,

You could try more conservative denoising settings, like increasing the number/order of motion covariates, or using more liberal outlier threshold, but if none of that appears to remove the residual QC-FC associations then consider also removing outlier subjects, as a few outlier subjects can also have a large effect on the QC-FC distributions. For example, you could re-create the QC-FC plots but this time removing different sets/levels of potential outlier subjects (e.g. when re-creating the plots unselect the 'all subjects' checkbox and either manually uncheck the potential outlier subjects or right-click and select any group 2nd-level covariate that you may have defined containing the subjects that you DO want to include in this plot) and see whether that helps improve the resulting quality control measures

Best
Alfonso

Originally posted by Hugo Morandini:

Great, thank you for that Alfonso!

Best,
Hugo