Fmriprep advices for multiband

Hi guys,

The data has been collected for a very well known task and I am very surprised that at the 2nd level analysis I do not see anything. I would like to know by any chance if someone has some advice about the best way to preprocess multiband data.

Am I removing too much signal during the preprocessing ? Am I not using the good preprocessing steps ?

My data is multiband (Simultaneous Multi-Slice) with :
TR = 400ms
Number of frames = 876
Number of slices in one volume := 48

With fmriprep I have used the following command line :
singularity run /share/PI/russpold/singularity_images/poldracklab_fmriprep_1.0.0-rc2-2017-08-12-d5eb3198f90a.img /scratch/PI/aetkin/redwan/framing/sourcedata/ /scratch/PI/aetkin/redwan/preprocessed/ participant --participant_label 01 --work-dir $LOCAL_SCRATCH --nthreads 16 --omp-nthreads 16 --no-freesurfer --write-graph

I have not used any slice time correction, according to this paper (

Then I have used 13 confounders at the 1st level analysis (6 moition parameters, FD, aComport)

Thank you very much,

We have successfully analyzed multiband data for various tasks using FMRIPREP with similar settings and using similar confound regressors on first level. The only exception is that we used --ignore slicetiming option.

What task is it? Are the HTML reports looking ok?

It may not be your whole problem, bu,t if you’re using aCompCor in 1.0.0-rc2, you should also include the “Cosine” confounds. CompCor has been run on high-pass-filtered BOLD series since 0.5.3.

@Chris - we would like to show the framing effect using ‘The financial decision-making task’ from De Martino. The HTML files look ok.

@Effigies - I can try to add “Cosine” regressor at the first level analysis but I even tried the first level analysis without aCompCor and FD. (I just tried with the 6 motions parameters) and the results were similar. The first level analysis looks ok (i see activation for all 20 subjects) but nothing at the second level.