Hello,
this is my first time trying to analyse fMRI data I got a bit stuck and, since my deadline is approaching, would greatly appreciate your help.
I've done the following:
- pre-processed data in FEAT -- did not click on Estimate (noise) from data, entered the 4D image, left the High pass filter cutoff at 90 s but did not turn high pass filtering on in Pre-stats. I used MCFLIRT motion correction, set slice timing correction to Interleaved, spatial smoothing to 6mm, checked BET, did not check MELODIC ICA data exploration. Under Registration: entered the brain extracted main structural image (Full search, BBR), registered to Standard space (the default option; Full search, 12 DOF + nonlinear, 10mm warp).
- ran ICA-AROMA, non-aggressive: python2.7 /Users/lucia/Desktop/fsl_scripts/ICA_AROMA.py -feat /Users/lucia/Desktop/fsl_new/data/OP242/10+.feat -out /Users/lucia/Desktop/fsl_new/data/OP242/10+.feat/ICA_AROMA/
- ran temporal filtering through a command (is there a way to run this only in the GUI?), hoping that would correspond to 90s: fslmaths /Users/lucia/Desktop/fsl_new/data/OP242/10+.feat/ICA_AROMA/denoised_func_data_nonaggr.nii.gz -bptf 18.0118373795 -1 /Users/lucia/Desktop/fsl_new/data/OP242/10+.feat/ICA_AROMA/temp_filtered.nii.gz
- tried to do Stats in FEAT (with FILM prewhitening, and Apply temporal filtering checked in the Full model setup). The Effect required/estimated noise changed to 0.0000 once I enter the temp_filtered image...
- got the following error in the report log, under Preprocessing Stage 2:
Preprocessing:Stage 2
/usr/local/fsl/bin/fslmaths prefiltered_func_data filtered_func_data
/usr/local/fsl/bin/fslmaths filtered_func_data -Tmin -bin mask -odt char
/usr/local/fsl/bin/fslstats filtered_func_data -k mask -R | awk '{ print }' -
ERROR:: Empty mask image
ERROR:: Empty mask image
ERROR:: Empty mask image
ERROR:: Empty mask image
ERROR:: Empty mask image
ERROR:: Empty mask image
ERROR:: Empty mask image
ERROR:: Empty mask image
ERROR:: Empty mask image
ERROR:: Empty mask image
ERROR:: Empty mask image
...
I tried this for 4 different participants, always got the same.
MY QUESTIONS:
1.) did I make a mistake at any point in my analysis?
2.) what does the error mean? I've searched the forum and found answers saying there's no signal left in my data after ICA-AROMA but that was when the error occurred later in the report.. Did I maybe just misplace a mask image?
3.) I don't believe that there's no signal left after ICA-AROMA, because:
If I run Full analysis, with high pass temporal filtering, in FEAT, even including motion outliers, I get results that look reasonable, although "Estimate noise from data" does give me really high (or really low) numbers (and I don't click on it). The denoised_func_data_nonaggr.nii.gz definitely looks better than the original one (clearly less movement). A LOT of the components get removed by ICA denoising but the time series looks very similar to the one after ICA-AROMA and temporal filtering. - The variance of the original vs. denoised vs. temp_filtered images is about the same (actually higher for the last one), only the temp_filtered one is centered around 0 for some reason (the other two are around 12000).
4.) what conclusion can I make from this? I'm considering leaving ICA-AROMA out and just using motion outliers but if this does mean there's no signal of interest left in my data after denoising, then maybe there's no point proceeding with the analysis at all.
Sorry this is so long, I tried to give any info that might be useful/help you help me figure out if/where I am making a mistake.
Thank you again! Really hope to get an answer soon as I don't really have anyone else to turn to.
Best regards,
Lucia
|