Hi!
AFNI version info (afni -ver):
Precompiled binary macos_13_ARM_clang: Jul 22 2024 (Version AFNI_24.2.01 'Macrinus')
I have been using MACAQUE_DEMO_REST to process the UC-Davis data repository that's available on PRIME-DRE. My goal is to perform a connectivity analysis. The functional and anatomical files are initially aligned. However, the FOV between the two files is different:
Anatomical File:
- Dimensions: (480, 512, 512)
- Voxel Size: (0.3, 0.30078125, 0.30078125)
- Field of View: (144.0, 154.0, 154.0)
Functional File:
- Dimensions: (200, 160, 36, 250)
- Voxel Size: (0.7, 0.7, 1.4, 1.6)
- Field of View: (140.0, 112.0, 50.4)
Likely because of the dimensions of T1, the animal warper fails:
Using 3DAutoBox command on T1 fixes the performance. However, I am then concerned about the performance of afni_proc. The resultant alignment between autoboxed T1 and intact functional seems slightly more misaligned than initially, with edges being most affected.
]
Typically, the edges are most affected, losing coverage that was previously there.
Here is my afni_proc command:
afni_proc.py \
-subj_id ${subj} \
-blocks tshift align tlrc volreg mask scale regress \
# -mask_apply anat \
-dsets ${dsets_epi} \
-copy_anat ${anat_cp} \
-anat_has_skull no \
-anat_uniform_method none \
-anat_follower GM anat ${sdir_aw}/GM_mask.nii.gz \
-anat_follower WM anat ${sdir_aw}/WM_mask.nii.gz \
-anat_follower CSF anat ${sdir_aw}/CSF_mask.nii.gz \
-anat_follower_erode WM \
-anat_follower_erode CSF \
-mask_segment_anat yes \
-mask_segment_erode yes \
-radial_correlate_blocks tcat volreg \
-radial_correlate_opts -sphere_rad 14 \
-tcat_remove_first_trs ${nt_rm} \
-volreg_align_to MIN_OUTLIER \
-volreg_align_e2a \
-volreg_tlrc_warp \
-volreg_warp_dxyz ${final_dxyz} \
-volreg_compute_tsnr yes \
-align_opts_aea -cost "${cost_a2e}" \
-cmass cmass -feature_size 0.5 ${aea_extra} \
-check_flip \
-tlrc_base ${ref_base} \
-tlrc_NL_warp \
-tlrc_NL_warped_dsets ${dsets_NL_warp} \
-regress_ROI_PC WM 3 \
-regress_ROI_PC CSF 3 \
-regress_motion_per_run \
-regress_apply_mot_types demean deriv \
-regress_censor_motion ${cen_motion} \
-regress_censor_outliers ${cen_outliers} \
-regress_est_blur_errts \
-regress_est_blur_epits \
-regress_run_clustsim no \
-html_review_style pythonic
EOF
Thus, my question is two fold: (1) is using 3DAutobox the optimal solution to deal with large dimension images with which animal warper fails to deal with and (2) if not, can modifications made to T1 affect later processing on afni proc, and how can I deal with this?
I would very much appreciate your feedback!
Sincerely,
Simona