What to do if a provided ICA mixing matrix doesn't result in likely BOLD components

[handwerkerd]

Summary

In the August 2023 dev call (#957) @BahmanTahayori commented that he had trouble tracking down a bug because it was not obvious that the ICA was re-run. The suggestion was to make a re-run of ICA more obvious in the outputs.

Additional Detail

When ICA is rerun the following warning message is added to the log file:

tedana/tedana/workflows/tedana.py

Lines 680 to 681 in f9daaa0

	if (n_restarts < maxrestart) and (n_likely_bold_comps == 0):
	LGR.warning("No BOLD components found. Re-attempting ICA.")

In writing this up, I'm realizing we might want an error rather than a more prominent warning. If a mixing matrix is calculated within tedana and doesn't converge or doesn't generate likely BOLD components, then reseeding and rerunning ICA an under-the-hood decision. We log when this happens, but it's not a major issue to highlight to users. The real issue here is when there is a user provided a mixing matrix and the output of tedana does not use that mixing matrix. It might make sense to end execution rather than output something that's different from what the user requested.

Next Steps

• Decide if this should be an error or a warning
  • Error: User-provided mixing matrix did not result in likely BOLD components. Rerun with a different user-provided mixing matrix or run ICA within tedana where ICA will be recalculated if no likely BOLD components are found
  • If we want a more prominent warning, what should be be?
    • An added line in the log WARNING: User-provided ICA mixing matrix did not result in likely BOLD components. tedana is recalculating ICA
    • A red line of text at the top of tedana_report.html

[tsalo]

I like the error idea.