Temporal ICA has not properly separated global fMRI signals: a comment on Glasser et al., 2018
Jonathan Power
Neuroimage 2019 Aug 15; 197: 650-651
Pubmed link
Figure (.pptx)
Below are files associated with the article "Temporal ICA has not properly separated global fMRI signals: a comment on Glasser et al., 2018". All movies are 1080p and will look best at full-screen resolutions. Movies stream from YouTube and are organized as playlists. Click on links next to the captions to download movies. Files to download are hosted on Dropbox Pro; if the links don't work our traffic has exceeded its 200GB/day limit and the links will be re-enabled the following day.
Movie 1 (.mov - 1.3 GB)
This movie shows the unprocessed physiology records for all subjects of the "900 subject" Human Connectome Project data release with 4 complete resting state fMRI scans and 4 accompanying complete sets of physiology traces. Pulse oximeter records are in orange and abdominal belt records are in blue. The total scan time is 14.4 minutes in all cases. The physiology data are acquired at 400 Hz. Units are arbitrary for both traces.
The questions an investigator must ask when viewing these records are:
1) Can I, using my eyes, identify peaks with good confidence?
2) Can I train a peak-finding algorithm to find the peaks I see?
-- and will it be worth the effort to correct the algorithm's mistakes (i.e., how mistake-laden will the results be)?
In my view, of the 760 subjects in the above movie (the subjects with 4 runs of fMRI data and 4 runs of physio data), 446 have physiology traces on which I would try to run and correct a peak-finding algorithm to obtain cardiac and respiratory parameters. My decisions on a subject by subject basis are here (.txt).
Software to help visualize and correct physiological records is here.