Hi,
I want to add more details about this issue. I found just 30 of my subjects appear this error when I processed all my 151 subjects. However, when I just dida single analysis choosing one subjects who was reported mismatched error, it works fine. And when I checked the CNN_X, indeed, I found the confounds are 205 row, while I only have 200 time points. Even I checked the subjects without error, I cannot understand how the confounds in CONN come from. Because the confounds are not same with they are in fmriprep's regressor.csv, eg,"QC_csf" is not the  "csf" in fmriprep's csv file. Very Appreciate if someone could explain what is the relationship between confounds in CONN and the original confounds in csv from fmriprep, since I found they are not the same. And I think maybe that is why I got 205 row confounds even though it is only 200 row in original csv file. I also attached my original regressor.csv from fmriprep.

Thanks!
Hongru