Dear FSL experts,
I would like to report what I think is an error in the output of mcflirt with specific inputs:
- bold timeserie with TR=1120ms and multi band factor of 5, 2mm isotropic resolution, 70 slices
- target image for the realignment: SBref image of the same serie, provided by the sequence of CMRR
for specific subjects who move very little
-> the motion parameters output by mcflirt give many zeros. Sometimes an entire column is full of zeros!
We first noticed these strange results while preprocessing our data with FMRIPREP (which uses mcflirt and SBref as reference when provided).
We also ran MRIQC on the same subject an the motion parameters illustrated by the FD time serie is very different than the one outputted by FMRIPREP. MRIQC gives the choice to use AFNI 3dvolreg or FSL-mcflirt, and in both case we get very similar results and no zero, on the contrary of FMRIPREP.
I tried many things, realign with SPM (reigning on first volume or SBref), mcflirt with realign on the mean, on the first volume, changing the cost function to mutualinfo, all the latter cases gave expected results, with no zero in the parameters files.
Only when using -cost leastsquares and SBref with mcflirt I got a par file with only zeros.
I described in detail this investigation here:
https://github.com/nipreps/fmriprep/issues/2360
What is your feeling about this? Would you like more information? For the moment the workaround I found with fmriprep was to not provide the SBref anymore, in that case FMRIPREP is looking as non steady state volumes to build its reference target for realignment and further registration to T1w image.
Thank you for your help!
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=FSL&A=1
This message was issued to members of www.jiscmail.ac.uk/FSL, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
|