Dear AFNI experts,
I am somewhat new to AFNI and had a basic question about the afni_proc.py output when using MEICA in tedana (such as Example 12c). I am familiar with the standalone tedana output, but was unsure if this folder contains just the intermediate files. Is the errts file in the main output folder the ultimate preprocessed file including all preprocessing steps PLUS the tedana optimally combined and denoised steps? I am attempting to use the preprocessed output for analysis in other software (such as FSL or CONN) and the results are bizarre when using seed-based correlation on the errts dataset and the dn_ts_OC dataset in the tedana folder, so not sure what I am doing wrong. The seeds from the QC file appear ok, to my eye (see attached PCC seed from QC report). Thank you in advance!
Sorry for missing this.
Yes, the final errts dataset (probably with tproject in the name, but that depends on the options applied in afni_proc.py) should have everything removed. You can play with that dataset in the InstaCorr demo to get a live view of how the seed-based correlations look.
If there are issues with other software, then there are 2 likely candidates:
-
Maybe they do not like the fact that the time series is already zero mean (though I am not sure why that would matter).
-
Maybe you did censoring, and they do not like having all-zero volumes in the residual time series (again, I am not sure why that would matter).
But you can easily look at correlations with seeds, even as you drag the seed location across the brain using the InstaCorr plugin. It seems like we do not have a training video in the “AFNI Academy”, but there is one from a course we did at MIT a few years back.
See https://cbmm.mit.edu/afni for the complete set.
InstaCorr is shown in class 31: “Resting State & InstaCorr: Part 2 of 2”. It might be worth watching the entire video for the details, but the smooth dragging of seed locations is of course done near the end. Note that if you use the errts time series as input for this, then turn off things like bandpassing and polort in the GUI.
- rick