fMRI Preprocessing on older adults

[raymondyangx]

Hi, AFNI experts,

I'm using AFNI proc.py and sswarper to preprocess an fMRI dataset. I found that the preprocessing quality is not very satisfactory, especially for older adults. I have attached an example of a participant who is 65 years old and displays noticeable atrophy and enlarged ventricles. The segmentation and normalization steps seem to fail in correctly extracting the lateral ventricles, and as a result, I observed strong yet peculiar activity in the CSF for my first-level results from a semantic task. I wonder if there are any special considerations I need to account for when preprocessing images of older adults, such as using an older adults atlas or adjusting specific parameters.

AFNI version info (afni -ver): Precompiled binary linux_openmp_64: Apr 9 2021 (Version AFNI_21.1.01 'Domitian')

I attached my afni proc_py script below.

SSwarper
/work/apps/AFNI/linux_openmp_64/@SSwarper -input /work/desai-lab/ABCAI/raw/sub-1024/anat/sub-1024_T1.nii - base MNI152_2009_template_SSW.nii.gz - subid sub-1024 - giant_move - deoblique - odir /work/desai-lab/ABCAI/raw/sub-1024/SSwarper


afni proc_py
#!/bin/tcsh
/work/apps/AFNI/linux_openmp_64/afni_proc.py -subj_id sub-1024 \
-scr_overwrite \
-copy_anat /work/desai-lab/ABCAI/raw/sub-1024/anat/sub-1024_T1.nii \
-anat_has_skull no \
-dsets /work/desai-lab/ABCAI/raw/sub-1024/func/sub-1024_fmri_task-Passage_raw.nii \
-blocks despike align tlrc volreg blur mask scale regress \
-radial_correlate_blocks tcat volreg \
-tcat_remove_first_trs 0 \
-align_opts_aea -cost lpc+ZZ \
-ginormous_move \
-resample off \
-deoblique on \
-check_flip \
-tlrc_base /work/apps/AFNI/linux_openmp_64/MNI152_2009_template_SSW.nii.gz \
-tlrc_NL_warp \
-tlrc_NL_warped_dsets /work/desai-lab/ABCAI/derivatives/sub-1024/sswarper/anatQQ.sub-1024.nii /work/desai-lab/ABCAI/derivatives/sub-1024/sswarper/anatQQ.sub-1024.aff12.1D /work/desai-lab/ABCAI/derivatives/sub-1024/sswarper/anatQQ.sub-1024_WARP.nii \
-volreg_align_to MIN_OUTLIER \
-volreg_align_e2a \
-volreg_tlrc_warp \
-mask_epi_anat yes \
-blur_size 6.0 \
-mask_segment_anat yes \
-regress_stim_times /work/desai-lab/ABCAI/raw/sub-1024/func/sub-1024_fmri_task-Passage_onset_passage.1D /work/desai-lab/ABCAI/raw/sub-1024/func/sub-1024_fmri_task-Passage_onset_foreign.1D \
-regress_stim_labels passage foreign \
-regress_opts_3dD -jobs 6 \
-gltsym 'SYM: passage[0] -foreign[0]' -glt_label 1 pass_NS \
-regress_basis WAV \
-regress_stim_types times \
-regress_ROI CSFe \
-regress_censor_motion 0.5 \
-regress_apply_mot_types demean deriv \
-regress_motion_per_run \
-regress_3dD_stop \
-regress_reml_exec \
-regress_compute_fitts \
-regress_make_ideal_sum sum_ideal.1D \
-regress_est_blur_epits \
-regress_est_blur_errts \
-regress_run_clustsim no \

[dglen]

Take a look at this alternative template here, mentioned in a previous thread:

template for elderly population?

Hello afni experts, I just have a quick question here: Does afni currently have a template that is based on an elderly population? Thank you very much in advance! Best Carolin

The template can be used within afni_proc.py by specifying the base template with -tlrc_base and -tlrc_NL_warp to compute a nonlinear warp to the template. @SSwarper won't be used in this case. The afni_proc.py generated script calls the auto_warp.py program to do the alignment instead.

[raymondyangx]

Thank you! I have a follow-up question. My dataset contains both young adults and older adults. Is it suggested to use different atlases for each group respectively, and then normalize to the same MNI space template?

Ray

[dglen]

Voxelwise comparisons are going to be problematic because of differences in atrophy and CSF. You can do an analysis of each group using their own appropriate template, and then compare the results in a combined template space, but you will have to carefully account for differences just from atrophy, as you have noticed. You can potentially mitigate this by using a measure of CSF/atrophy as a covariate. An ROI analysis using ROIs from transformed atlases or manually drawn regions, instead of a voxelwise analysis, may help.

Atrophy measure suggestion

This is a fairly complicated topic really. Regarding the specific questions of atrophy measures, there are a few possibilities: CSF probability. Using probabilistic segmentation from a program like 3dSeg or similar, you can generate probabilities of different tissue classes. Gray matter probability is another option here. Deformation to a normal template with Jacobian to get a compression/expansion type metric. https://afni.nimh.nih.gov/afni/community/board/read.php?1,165536,165558#msg-16…

[raymondyangx]

Hi, thank you so much for the suggestion. We'd like to regress out the signal in CSF using "regress_ROI". The CSF and CSFe masks were created for every subject. We checked the masks and for some participants the masks do not contain the lateral ventricles. We thought this problem was because of the structural change of the older people and thus we should turn to use a age-specific atlas. But after more exploration these days, we found that the CSF masks for several young adults are also not generated correctly because the lateral ventricles were also missing (I attached an uneroded CSF mask of a young adult for an example). I checked the preprocessed data and the output files from sswarper. It seems that the registration and normalization were fine. Do you have any suggestions on how to segment the CSF correctly?

Xuan

[pmolfese]

Would be interesting to see the other masks (GM, WM, CSFe)...

Some ideas... but first: For comparison, I'd try running one of the troublesome subjects through the Freesurfer pipeline and then @SUMA_Make_Spec_FS to see how the mask there looks in comparison.

What percentage of subjects have a bad CSF mask?