Meeting 28-02-2022

[gjheij]

Meeting notes

We mostly discussed the right approach concerning our preprocessing pipeline. Ideally, we'd like something like this:

Preprocessing: 1) NORDIC (default) 2) Detrend (high-pass filter e.g., DCT) 3) RETROICOR/aCompCor

For 3), we contacted Phillips due to corrupt SCANPHYSLOG-files. The traces looked adequate on the console, but the .log-files are practically empty. We eagerly await Phillips' response.

As far as aCompCor is concerned, we know how to get the tissue classifications in the correct space using this. I guess the idea is then to create one big dataframe including physiology and PCA'd timecourses of white matter/CSF voxels and regress that out. We should inspect the power spectra after as well as the timecourses themselves to see if they're suited for pRF-fitting.

pRF-fits:

• Add baseline of 4 runs (EDIT after meeting: this doesn't seem to help much with the first sweep, so I kept it out for now).
• Try fit with fixed x/y; this could be a last resort if independent fitting fails to produce pRF-fits close to the target parameters.
• 2nd order polynomial for size/depth plot; minor detail, but this would allow for a U-shaped'ish graph as produced by Alessio

Exp:

• Remove "waiting for scanner"

To do:

1) pRF-fitting of ses-4-data. Verify that sparse design matrix results in more sensible parameters (also fit HRF from predictions) 2) Implement aCompCor in pipeline 3) ..

[gjheij]

See subsequent issues for updates on pRF-fits and aCompCor