afni_proc.py missing sub brick data

Hi experts,

After using afni_proc.py to preprocess some resting-state fMRI data, there are always some sub-bricks become zeros in the results. I tried several times and got the same results every run. For example, one of the 3dinfo infomation of one subject:


3dinfo -verb errts.S126.tproject+tlrc.


– At sub-brick #68 ‘S126.rs.02aptopu[68]’ datum type is float: -48.3584 to 17.7858
– At sub-brick #69 ‘S126.rs.02aptopu[69]’ datum type is float: 0 to 0
– At sub-brick #70 ‘S126.rs.02aptopu[70]’ datum type is float: 0 to 0
– At sub-brick #71 ‘S126.rs.02aptopu[71]’ datum type is float: -19.4341 to 23.5937

my code


afni_proc.py                                                 \
      -subj_id "$i"                                        \
      -dsets "$i".rs.02aptopup.nii.gz                            \
      -copy_anat "$i".T1.padded180.nii                                     \
      -blocks despike tshift align tlrc volreg blur mask scale regress  \
      -tcat_remove_first_trs 0                                  \
      -volreg_align_to MIN_OUTLIER                              \
      -volreg_align_e2a \
      -volreg_tlrc_warp \
      -mask_apply anat                                             \
      -mask_segment_anat yes                                      \
      -mask_segment_erode yes                                     \
      -blur_size 6                                              \
      -regress_censor_motion 0.5                               \
      -regress_censor_outliers 0.05                             \
      -regress_bandpass 0.01 0.1                                \
      -regress_apply_mot_types demean deriv                     \
      -regress_est_blur_epits                                   \
      -regress_est_blur_errts

Could you provide some advice? Thank you!

Xiyue

Hi Xiyue,

Any volume that was censored from the linear regression (due to motion parameters or outliers in this case) will be zero in the errts time series. Those zero volumes will not affect correlations among voxels in that residual dataset.

Note that band pass filtering will probably eat up the majority of your degrees of freedom.
See the “RESTING STATE NOTE” from the output of afni_proc.py -help: for details.

  • rick