Hi @Jinyu_WANG, and welcome to neurostars!
Looks like a freesurfer command minc_modify_header was excluded from the fmriprep container. However, mri_total should only be running on images that were poor quality and thus had to go through some extra recon-all steps. You could run freesurfer externally and then pass that into fmriprep with the —fs-subjects-dir argument, but I would see first if that subject’s data is even usable.
Best,
Steven