I used tedana combine/denoise the multi echo data pre-processed using fMRIprep. Why do some parts of the brain are missing in TEDANA output? TEDANA command used ,
tedana -d sub-13023_ses-1_task-Insight1_run-01_echo-1_space-MNI152NLin2009cAsym_desc-preproc_bold.nii.gz sub-13023_ses-1_task-Insight1_run-01_echo-2_space-MNI152NLin2009cAsym_desc-preproc_bold.nii.gz sub-13023_ses-1_task-Insight1_run-01_echo-3_space-MNI152NLin2009cAsym_desc-preproc_bold.nii.gz -e 10.8 28.68 46.56 --out-dir insight1
Here are the three echos(after fMRIPrep pre-processed)
TEDNA output : (upload://m1Ii8beCTdDwCoUD2W2ztY0FyKy.png)
Here is the TEDANA output
Hi @sameera2004 –
Unfortunately, we don’t recommend this pipe-lining of the two methods – the distortion correction applied in fMRIPrep invalidates the calculation of the T2* map that underlies tedana’s optimal combination and denoising procedures. Instead, we recommend using files from the fMRIPrep working directory. Please see this thread for more details: Combining fmriprep with ME-ICA from tedana
I would also add that from my experience feeding in minimally processed images to tedana, as @emdupre, suggests can also result in tedana cutting off some of brain (mostly subcortical). One way I have circumvented this issue is by feeding in an explicit mask to tedana using -m.
Hi @emdupre and @gspitz,
Thank you so much for your replies. Yes it worked when I run TEDANA from the fMRIprep working directory and providing the mask.(–mask)
Happy to hear this worked for you ! Thanks for following up, @sameera2004
I have another question regarding these TEDANA outputs (dn_ts_OC.nii).
Do I need to run one more time fMRIprep for these TEDANA output to perform regression with the denoised data? Is there a easy way to do this? Because I need to prepare them again in BIDS format etc to run fMRIprep?
So it’s basically adapting the fmriprep t2smap command for tedana, right?
I see that this command is using skull stripped files while the python script at https://github.com/ME-ICA/tedana-reliability-analysis/blob/master/collect_fmriprep.ipynb refers to non-skullstripped files in a different location. As the images in both locations seem to have the same values within the mask area, I assume that there are no other transformations applied.
Furthermore, it seems that the input will be masked anyway using Nilearn’s compute_epi_mask.