Fatal error at @radial_correlate step in afni_proc.py

AFNI Message Board
1

Hellooo. I'm doing some QC checks on pilot ME data and appear to be having the same issue as this poster: Errors while running correlation in afni_proc.py

AFNI version info (afni -ver): 25.0.00

Here's the script:

afni_proc.py                                                                 \
    -subj_id                  test                                            \
    -blocks                   despike tshift align volreg mask combine blur scale regress  \
    -radial_correlate_blocks  tcat volreg                            \
    -copy_anat                uni_denoised_brain.nii.gz                      \
    -anat_has_skull	      no					     \
    -anat_follower	      anat_w_skull anat uni_denoised.nii	     \
    -dsets_me_run             e1.nii e2.nii e3.nii                           \
    -echo_times               11 26.45 41.9                                  \
    -combine_method           m_tedana                                             \
    -reg_echo                 1                                              \
    -tcat_remove_first_trs    2                                              \
    -tshift_interp            -wsinc9                                        \
    -align_unifize_epi         local                                 \
    -mask_epi_anat             yes                                   \
    -align_opts_aea            -cost lpc+ZZ                          \
                               -giant_move                           \
                               -check_flip                           \
    -volreg_align_to          MIN_OUTLIER                                    \
    -volreg_align_e2a                                                \
    -volreg_warp_final_interp  wsinc5                                \
    -volreg_compute_tsnr      yes                                            \
    -blur_size                4                                              \
    -regress_motion_per_run                                                  \
    -regress_apply_mot_types  demean deriv                                   \
    -html_review_style        pythonic

Here are the last few lines with the error:

-- running correlation on dataset pb00.test.r01.e01.tcat+orig.HEAD ...

++ merge blur: rad 20, FWHM 40
++ 3dmerge: AFNI version=AFNI_25.0.00 (Jan 14 2025) [64-bit]
*** cannot open dataset det.r01+orig
Program 3dmerge
3dmerge: edit and combine 3D datasets, by RW Cox
++ 3dTcorrelate: AFNI version=AFNI_25.0.00 (Jan 14 2025) [64-bit]
** FATAL ERROR: Can't open input dataset sphere.mean.20.r01+orig
** Program compile date = Jan 14 2025
** missing correlation result radcor.20.r01.corr+orig.HEAD

From comparing the proc script with its output the error seems to stem from the QC: compute correlations with spherical ~averages step with @radial_correlate.

Any help would be appreciated! Thank you!

2

Hello,

Your command is working for me. There might be something odd about the datasets.

Would you please send me email (I will send you a message) with the output.proc.* text file?

Also, what is the output from:

3dhistog -nbin 10 e1.nii

Thanks,
-rick

3

Thanks for the details. There isn't anything too special in the histogram, aside from converting unsigned shorts to floats. But the real problem is shown in the full output, just above what you posted here. It is a message from the operating system:

Killed

That error most typically happens when you do not have sufficient RAM for the processing being done. And note the fact that since these are converted from unsigned short to float 32, that will double the size of the data.

But since this is only a single run, it is surprising that such a simple operation as detrending runs out of RAM. How much RAM do you have and how big is the dataset? Consider:

3dinfo -n4 e1.nii

Thanks,
-rick