Look for inefficiencies in the interaction between unmasking and spatial clustering

[handwerkerd]

Summary

As part of examining #490, I noticed that utils.unmask was called 400 times through the tedana workflow. 352 of those 400 times were following the logged line:
"INFO:tedana.metrics.kundu_fit:Performing spatial clustering of components" My data has 323 volumes.

Next Steps

• This section of the code should be examined to see if there is any way to call unmask less often or to make unmasking more efficient in this situation.
• unmask should be profiled with different dataset sizes to get a sense of the computational costs of these repeated calls (i.e. is this inefficient enough to make optimization worth the effort). Profiling unmask might be a good first issue for someone.
• This inefficiency relates to my general concern that always moving files around in an img format might end up compounding inefficiencies like this.