Issue when running ICA-AROMA

[KarisColyerPatel]

Describe the bug
When running the confound regression step (with the inclusion of ICA-AROMA) the output is empty files (called empty.nii.gz!). If ICA-AROMA is not run the output looks fine!

Describe RABIES call
singularity run -B /home/c12630594/Desktop/input_BIDS_full:/input_BIDS_full:ro -B /home/c12630594/Desktop/preprocess_outputs_ratsremoved:/preprocess_outputs_ratsremoved/ -B /home/c12630594/Desktop/confound_corrections_outputs5_withica:/confound_corrections_outputs5_withica/ /home/c12630594/rabies.sif -p MultiProc confound_correction /preprocess_outputs_ratsremoved/ /confound_correction_outputs5_withica/ --conf_list WM_signal CSF_signal mot_6 --smoothing_filter 0.5 --TR 0.801 --ica_aroma apply=true,dim=30,random_seed=1 --highpass 0.01

Attach log file
Attach to your issue the .log files present in the output folder. (e.g. rabies_out/rabies_preprocess.log)
rabies_confound_correction.log

[Gab-D-G]

Hi Karis,
I am not sure what has happened, but it is not clear if the pipeline completed, as the last step in the log looks like it was still running for one scan. Is this an issue only with one scan? Can you print here the content of the folder you are inspecting?

[KarisColyerPatel]

Hey Gabriel,

I re-ran to completion (log attached below). Unfortunately, it's an issue with all scans!

rabies_confound_correction.log

The output folders look as follows:

Confound_correction_datasink/cleaned_timeseries/_split_name_sub10_r1_a2_d2_b1_bold/empty.nii.gz (this is the same for all rats)

Confound_correction_datasink/aroma_out/_split_name_sub10_r1_a2_d2_b1_bold/empty.nii.gz (this is the same for all rats)

Confound_correction_datasink/plot_CR_overfit/_split_name_sub10_r1_a2_d2_b1_bold/empty.nii.gz (this is the same for all rats)

[Gab-D-G]

Hi Karis,

the issue you are encountering happens inside the ICA-AROMA package, which unfortunately does not generate proper error reports (except with the most recent version 0.5.1). I would suspect that MELODIC didn't run for some reason, and the script kept running without throwing a clear error. The issue could arise from inappropriate formats of the input files, which include the CSF mask and motion parameters together with the functional timeseries. If the spatial or temporal dimensions are unusual for your temporal files, this could generate issues. Also, if there were registration errors during preprocessing, these errors could propagate. These issues should be better mitigated in 0.5.1, otherwise my help remains limited without further insights.

Best,
Gabriel