Missing ICA-AROMA denoising outputs

Hello - I have found that, for a subset of runs (2 / 25 runs, all from the same subject) that I am analyzing, ICA-AROMA runs without any obvious errors… but does not produce the default denoised “_nonaggr.nii.gz” file (or any denoised file). All other files are present and there is no error produced in the text output to the terminal. I am reaching out to the experts here because I am running out of ideas as to why this happened.

The only thing that distinguishes the two runs that failed from the others that ran without issue, as far as I can tell, is that the number of components identified is greater (>400 vs. ~100-150 on average). It is not the case, however, that all the components were classified as motion-related, like I have seen others post about in this forum (i.e., there were still a reasonable number of components identified as signal). I have also visually inspected all the components and they look fine and the classifications seem appropriate. I don’t think this is a memory issue either, I allocated more than enough memory on the cluster I running this analysis on.

Hoping that someone has some insight into this issue, thank you for your time!

In case it is useful for others - I was able to resolve this issue by explicitly limiting the dimensionality of the ICA decomposition (150 components), as suggested in this other thread (Testing different noise models output from fmriprep - unclear results with ICA-AROMA) for high-volume MB data (which is the case here for my data).

1 Like

Glad you sorted this out. Thanks for posting the solution!