Hi Robert,

Thank you very much for the detailed description. This was actually a bug in the code which resulted in projects needing to be manually saved before running QA plots when using parallelization options. In general, as you expected, there is no need to click on 'Done' in the Denoising tab for the QA denoising plots to properly read the current information from the Denoising tab in your project (but the bug above was causing the plot creation function to be reading the incorrect/old values from the conn project file instead). 

To avoid this issue you may save your CONN project before running QA plots in parallel. Alternatively, to actually fix this issue you may update to the latest development release (available in github) where I have uploaded a patch for this. Please let me know if you still run into any issues

Hope this helps and thanks again for the detailed explanation!

Alfonso

Originally posted by Robert Claar:

I am using CONN 22a and need to update to the most recent version to fix another issue I posted about that you recentlly answered, so perhaps the issue I describe more below will be solved from this as well.

I am specifically looking at the "QA Denoising: Distribution of functional connectivity values" plot. After I click "Done" on the "Setup" tab and CONN finishes setting up my data and brings me to the "Denoising (1st-level)" tab the default confounds present are:

White Matter (5P)

CSF (5P)

realignment (12P)

scrubbing (variable num of parameters)

Effect of task condition #1

Effect of task condition #2

and the band-pass filter is set to [0.008 0.09] Hz.

Before changing anything I click through a few of my subjects and look at their distributions of connectivity values. Then, based on my task-design and TR, I do not want to low-pass filter my data so I change the filter bounds to [0.025 Inf]. I then play around with different sets of confounds checking a few subjects connectivity distributions each time and find a set that seems to do a good job at denoising my data based on the 10 or so subjects I looked at so I want to make a QA plot to view all of the distributions of connectivity values together.

I click QA plots, create a new report, and select "QA Denoising: Distribution of functional connectivity values" and a couple other plots. I run this in parallel submitting as many jobs as I have subjects (a side note, for HPC users, it would be helpful to have the maximum possible number of jobs shown when creating QA plots as is shown when preprocessing, denoising, etc in parallel).

When the jobs complete and I go to look at the distributions of connectivity values for all my subjects, the distributions do not correspond to those generated by my selected confounds, but rather are the same distributions I saw when first landing on the "Denoising (1st-level)" tab and further when I minimize the QA report window and go back to the "Denoising (1st-level)" tab, my confounds have been reset to the default confounds that were present at the beginning.

I eventually clicked through and looked at each subjects distribution of connectivity values individually in the GUI and then ran the denoising. Afterwards as a test I recreated the QA plots and they now reflect my chosen set of confounds. Do I always need to click "Done" and actually denoise the data before creating the QA report and the corresponding plots? I figured I wouldn't need to since the individual distributions of connectivity values plots are updated automatically as you select new confounds in the "Denoising (1st-level)" tab.