hmolla
July 20, 2022, 4:40pm
1
Hello,
I’m fairly new to imaging analysis and I am trying to develop a new proc script from this one I was given. Right now, this script has 6 contrasts for a reward task, 2 of which are:
wins($1 and $5) vs. no wins ($0)
I’d like to separate out each of the magnitudes so there are:
What should be changed/added here?
-regress_stim_times \
$stim_dir/loss_0_ant.txt \
$stim_dir/loss_ant.txt \
$stim_dir/win_0_ant.txt \
$stim_dir/win_ant.txt \
$stim_dir/fix.txt \
$stim_dir/win_fb_hit.txt \
$stim_dir/win_fb_nhit.txt \
$stim_dir/loss_fb_hit.txt \
$stim_dir/loss_fb_nhit.txt \
-regress_stim_labels \
L0ANT \
LANT \
W0ANT \
WANT \
FIX \
WFB_HIT \
WFB_NHIT \
LFB_HIT \
LFB_NHIT \
-regress_basis_multi \
'GAM' \
'GAM' \
'GAM' \
'GAM' \
'GAM' \
'GAM' \
'GAM' \
'GAM' \
'GAM' \
-regress_stim_types \
times \
times \
times \
times \
times \
times \
times \
times \
times \
-regress_censor_motion 0.3 \
-regress_apply_mot_types demean deriv \
-regress_motion_per_run \
[b]-regress_opts_3dD \
-num_glt 6 \
-gltsym 'SYM: 1.0*WANT -1.0*W0ANT' -glt_label 1 ANT_WIN_VS_NOWIN \
-gltsym 'SYM: 1.0*LANT -1.0*L0ANT' -glt_label 2 ANT_LOSS_VS_NOLOSS\
-gltsym 'SYM: 1.0*WANT -1.0*LANT' -glt_label 3 ANT_WIN_VS_LOSS \
-gltsym 'SYM: 0.5*WANT 0.5*LANT -0.5*W0ANT -0.5*L0ANT' -glt_label 4 ANT_W15L15_VS_W0L0 \
-gltsym 'SYM: WFB_HIT -WFB_NHIT' -glt_label 5 WINHIT_VS_MISS \
-gltsym 'SYM: LFB_HIT -LFB_NHIT' -glt_label 6 LOSSHIT_VS_MISS \[/b]
-regress_anaticor_fast \
-regress_reml_exec \
-regress_make_ideal_sum sum_ideal.1D \
-regress_est_blur_epits \
-regress_est_blur_errts \
-regress_run_clustsim no
rickr
July 20, 2022, 5:02pm
2
Hello,
It looks like 5 of the 6 extra symbolic contrasts (SYM) are simple pairs as you describe, with only contrast 4 being a difference in means.
Thanks,
hmolla
July 20, 2022, 5:54pm
3
Srry I should have mentioned in the original post that ANT_WIN_VS_NOWIN and ANT_LOSS_VS_NOLOSS correspond to Win ($1 and $5) vs No Win ($0) and Loss ($1 and $5) vs No Loss ($0), respectively. Each magnitude separated is not listed here. I’d have to include those to the script. In fact, I tried that using the following script, and did not get the same output. Thoughts on how to fix this? Thanks.
afni_proc.py -subj_id $subj WANT -1.0 W0ANT’ -glt_label 1 ANT_WIN_VS_NOWIN LANT -1.0 L0ANT’ -glt_label 2 ANT_LOSS_VS_NOLOSSWANT -1.0 LANT’ -glt_label 3 ANT_WIN_VS_LOSS WANT 0.5 LANT -0.5W0ANT -0.5 L0ANT’ -glt_label 4 ANT_W15L15_VS_W0L0 W1ANT -1.0 W0ANT’ -glt_label 7 ANT_WIN1_VS_NOWIN W5ANT -1.0 W0ANT’ -glt_label 8 ANT_WIN5_VS_NOWIN L1ANT -1.0 L0ANT’ -glt_label 9 ANT_LOSS1_VS_NOLOSS L5ANT -1.0 L0ANT’ -glt_label 10 ANT_LOSS5_VS_NOLOSS
Hi-
I would just like to reformat the afni_proc.py command a bit to help for readability (and ease for adding/editing options). This might help viewing a bit:
afni_proc.py \
-subj_id $subj \
-script proc.$subj \
-scr_overwrite \
-blocks despike tshift align tlrc volreg blur mask \
scale regress \
-copy_anat $anat_dir/anatSS.${subj}.nii \
-anat_has_skull no \
-tcat_remove_first_trs 0 \
-tshift_opts_ts -tpattern seqplus \
-dsets $epi_dir/${subj}.MID.epi+orig. \
-volreg_align_to MIN_OUTLIER \
-tlrc_base $top_dir/nlWarp/MNI152_2009_template_SSW.nii.gz \
-volreg_align_e2a \
-volreg_tlrc_warp \
-tlrc_NL_warp \
-tlrc_NL_warped_dsets $anat_dir/anatQQ.${subj}.nii \
$anat_dir/anatQQ.${subj}.aff12.1D \
$anat_dir/anatQQ.${subj}_WARP.nii \
-blur_size 5.0 \
-mask_epi_anat yes \
-regress_stim_times $stim_dir/loss_0_ant.txt \
$stim_dir/loss_1_ant.txt \
$stim_dir/loss_5_ant.txt \
$stim_dir/loss_ant.txt $stim_dir/win_0_ant.txt \
$stim_dir/win_1_ant.txt \
$stim_dir/win_5_ant.txt $stim_dir/win_ant.txt \
$stim_dir/fix.txt $stim_dir/win_fb_hit.txt \
$stim_dir/win_fb_nhit.txt \
$stim_dir/loss_fb_hit.txt \
$stim_dir/loss_fb_nhit.txt \
-regress_stim_labels L0ANT L1ANT L5ANT LANT W0ANT W1ANT W5ANT WANT \
FIX WFB_HIT WFB_NHIT LFB_HIT LFB_NHIT \
-regress_basis_multi 'GAM' 'GAM' 'GAM' 'GAM' 'GAM' 'GAM' 'GAM' \
'GAM' 'GAM' 'GAM' 'GAM' 'GAM' 'GAM' \
-regress_stim_types times times times times times times times \
times times times times times times \
-regress_censor_motion 0.3 \
-regress_apply_mot_types demean deriv \
-regress_motion_per_run \
-regress_opts_3dD -num_glt 10 \
-gltsym 'SYM: 1.0*WANT -1.0*W0ANT' \
-glt_label 1 ANT_WIN_VS_NOWIN \
-gltsym 'SYM: 1.0*LANT -1.0*L0ANT' \
-glt_label 2 ANT_LOSS_VS_NOLOSS-gltsym \
'SYM: 1.0*WANT -1.0*LANT' \
-glt_label 3 ANT_WIN_VS_LOSS \
-gltsym \
'SYM: 0.5*WANT 0.5*LANT -0.5*W0ANT -0.5*L0ANT' \
-glt_label 4 ANT_W15L15_VS_W0L0 \
-gltsym 'SYM: WFB_HIT -WFB_NHIT' \
-glt_label 5 WINHIT_VS_MISS \
-gltsym 'SYM: LFB_HIT -LFB_NHIT' \
-glt_label 6 LOSSHIT_VS_MISS \
-gltsym 'SYM: 1.0*W1ANT -1.0*W0ANT' \
-glt_label 7 ANT_WIN1_VS_NOWIN \
-gltsym 'SYM: 1.0*W5ANT -1.0*W0ANT' \
-glt_label 8 ANT_WIN5_VS_NOWIN \
-gltsym 'SYM: 1.0*L1ANT -1.0*L0ANT' \
-glt_label 9 ANT_LOSS1_VS_NOLOSS \
-gltsym 'SYM: 1.0*L5ANT -1.0*L0ANT' \
-glt_label 10 ANT_LOSS5_VS_NOLOSS \
-regress_anaticor_fast \
-regress_reml_exec \
-regress_make_ideal_sum sum_ideal.1D \
-regress_est_blur_epits \
-regress_est_blur_errts \
-regress_run_clustsim no
With that being done, I might suggest:
reordering slightly so all sub-options are grouped a bit (for ease of reading)
adding some “-align_* …” options, to help EPI-anatomical alignment and QC
-align_opts_aea -cost lpc+ZZ \
-giant_move \
-check_flip \
add in some more options for QC viewing, including TSNR estimation after volreg and the radcor block:
-volreg_compute_tsnr yes \
-radial_correlate_blocks tcat volreg \
add in Python QC HTML creation (which is now the default, but I am not sure what version of AFNI you have):
afni_proc.py \
-subj_id $subj \
-script proc.$subj \
-scr_overwrite \
-blocks despike tshift align tlrc volreg blur mask \
scale regress \
-copy_anat $anat_dir/anatSS.${subj}.nii \
-anat_has_skull no \
-dsets $epi_dir/${subj}.MID.epi+orig. \
-tcat_remove_first_trs 0 \
-tshift_opts_ts -tpattern seqplus \
-radial_correlate_blocks tcat volreg \
-align_opts_aea -cost lpc+ZZ \
-giant_move \
-check_flip \
-volreg_align_to MIN_OUTLIER \
-volreg_align_e2a \
-volreg_tlrc_warp \
-volreg_compute_tsnr yes \
-tlrc_NL_warp \
-tlrc_base $top_dir/nlWarp/MNI152_2009_template_SSW.nii.gz \
-tlrc_NL_warped_dsets $anat_dir/anatQQ.${subj}.nii \
$anat_dir/anatQQ.${subj}.aff12.1D \
$anat_dir/anatQQ.${subj}_WARP.nii \
-blur_size 5.0 \
-mask_epi_anat yes \
-regress_stim_times $stim_dir/loss_0_ant.txt \
$stim_dir/loss_1_ant.txt \
$stim_dir/loss_5_ant.txt \
$stim_dir/loss_ant.txt $stim_dir/win_0_ant.txt \
$stim_dir/win_1_ant.txt \
$stim_dir/win_5_ant.txt $stim_dir/win_ant.txt \
$stim_dir/fix.txt $stim_dir/win_fb_hit.txt \
$stim_dir/win_fb_nhit.txt \
$stim_dir/loss_fb_hit.txt \
$stim_dir/loss_fb_nhit.txt \
-regress_stim_labels L0ANT L1ANT L5ANT LANT W0ANT W1ANT W5ANT WANT \
FIX WFB_HIT WFB_NHIT LFB_HIT LFB_NHIT \
-regress_basis_multi 'GAM' 'GAM' 'GAM' 'GAM' 'GAM' 'GAM' 'GAM' \
'GAM' 'GAM' 'GAM' 'GAM' 'GAM' 'GAM' \
-regress_stim_types times times times times times times times \
times times times times times times \
-regress_censor_motion 0.3 \
-regress_apply_mot_types demean deriv \
-regress_motion_per_run \
-regress_opts_3dD -num_glt 10 \
-gltsym 'SYM: 1.0*WANT -1.0*W0ANT' \
-glt_label 1 ANT_WIN_VS_NOWIN \
-gltsym 'SYM: 1.0*LANT -1.0*L0ANT' \
-glt_label 2 ANT_LOSS_VS_NOLOSS-gltsym \
'SYM: 1.0*WANT -1.0*LANT' \
-glt_label 3 ANT_WIN_VS_LOSS \
-gltsym \
'SYM: 0.5*WANT 0.5*LANT -0.5*W0ANT -0.5*L0ANT' \
-glt_label 4 ANT_W15L15_VS_W0L0 \
-gltsym 'SYM: WFB_HIT -WFB_NHIT' \
-glt_label 5 WINHIT_VS_MISS \
-gltsym 'SYM: LFB_HIT -LFB_NHIT' \
-glt_label 6 LOSSHIT_VS_MISS \
-gltsym 'SYM: 1.0*W1ANT -1.0*W0ANT' \
-glt_label 7 ANT_WIN1_VS_NOWIN \
-gltsym 'SYM: 1.0*W5ANT -1.0*W0ANT' \
-glt_label 8 ANT_WIN5_VS_NOWIN \
-gltsym 'SYM: 1.0*L1ANT -1.0*L0ANT' \
-glt_label 9 ANT_LOSS1_VS_NOLOSS \
-gltsym 'SYM: 1.0*L5ANT -1.0*L0ANT' \
-glt_label 10 ANT_LOSS5_VS_NOLOSS \
-regress_anaticor_fast \
-regress_reml_exec \
-regress_make_ideal_sum sum_ideal.1D \
-regress_est_blur_epits \
-regress_est_blur_errts \
-regress_run_clustsim no \
-html_review_style pythonic
There may be other suggestions for task modeling and things, too.
–pt
Oh, and maybe add in one more censoring criterion:
-regress_censor_outliers 0.05 \
… which would censor EPI volumes that have more than 5% of the brainmask filled with outlier values at a given time point?
–pt
rickr
July 22, 2022, 1:52am
6
Hello,
Sorry, but I am still a bit confused.
You say that the ANT_WIN_VS_NOWIN contrast corresponds to Win ($1 and $5) vs No Win ($0). However ANT_WIN_VS_NOWIN seems to be defined as a simple contrast of ‘WANT -W0ANT’, suggesting WANT is itself Win ($1 and $5), and No Win $0 is -W0ANT (that part seems reasonble).
Would you be able to clarify ‘Win’ more? Exactly what is Win $1? What is Win $5?
Thanks,