Hi Rick,

So the data is archival and unfortunatley a rapid event related design with randomly interspersed conditions (ie. no convenient blocking); obviously not ideal for FC. Therefore we did 3dDeconolve -errts independently for each condition. For condition A: regressing out B + C, for B: regressing out C + A etc and so forth. Then within each version of the data we cut out task events of no interest (leaving single rows of NaN to break them up), creating "simulated blocks" of just our desired condition with -in theory- the residual effects of the recovery and refractory periods of the surrounding non-interest task events removed. 3dMaskAv generated time series from each of these newly agregated task "blocks" to create pearsons correlations for each condition. This would allow us to run ttests of ROI-to-ROI FC coeffs between the different task conditions. That - in a larger nutshell - was the rationale for regressing only tasks of no interest. There was also the added benefit that these new "blocks" made the data suitable for entering into an effective connectivity modeling program we used to assess if probable pathways differed by condition. If there is anything grossly wrong with this way of -errts + data parsing which we didn't consider, please let me know.

I couldn't find any materials or AFNI -help files on how to insert bandpassing directly into my existant 3dDeconvolve script. Is there a -option, like "-bandpass_freq" which I am missing somewhere? Also If I do go this route I would do scaling beforehand and concat runs after, correct? I also need to admit to my naivety and confusion about what your suggestion of using 3dTproject with the resulting 3dDecon X-matrix functionlly accomplishes.

I read up on Afni_proc.py and wrote a script, as you suggested, for my continued education and as a backup option. My apologies, I didn't realize how adaptable afni_proc is and that I could use it for just portions of my analysis. I see now why you suggested it so strongly. My attempt to edit Example 9 from -helpfile to accomplish 3dDeconvolve+Bandpassing+censoring+motion+scaling all simultaneously is below. I would most likely concatentate runs afterwards correct?

I am obviously new to python and this particular script so feel free to correct me and edit any slip ups.

--------------------------------------------------------------------------------------------------------------------------------------------------
#!/bin/tcsh

set subjects = ( 2H 5U etc...)
set subjectsdir = /Volumes/...
set censorfile = /Volumes/....
set stimdir = /Volumes/.....

cd ${subjectsdir}

foreach subj ( 2H 5U etc...)

cd ${subj}

foreach sess ( s4 s5 s6 s7 s8 s9 )

afni_proc.py -dsets ${subj}_$(sess)_vr_zp4_cardnudg_e2a_FWHM+tlrc. \
-out_dir /Volumes/.../ICN_Runs \
-force_TR 2.5 \
-regress_censor_extern $censorfile/${subj}_$(sess)_autocensor_AC3VC_pre-moco.1D \
-regress_polort 3
-regress_apply_mask ${subj}_s4_vr_zp4_cardnudg_e2a_amsk+tlrc. \
-regress_basis 'TENT(2.5,17.5,7)' \
-regress_local_times \
-regress_stim_times 1 $stimdir/${subj}_Strong_stim4b.1D -regress_stim_label 1 'Strg' \
-regress_stim_times 2 $stimdir/${subj}_Weak_stim4b.1D -regress_stim_label 2 'Weak' \
-regress_stim_file 3 $stimdir/Motion_Pre-moco/${subj}_${sess}_motion.1D'[0]' -regress_stim_label 3 'roll' -regress_stim_base 3 \
-regress_stim_file 4 $stimdir/Motion_Pre-moco/${subj}_${sess}_motion.1D'[1]' -regress_stim_label 4 'pitch' -regress_stim_base 4 \
-regress_stim_file 5 $stimdir/Motion_Pre-moco/${subj}_${sess}_motion.1D'[2]' -regress_stim_label 5 'yaw' -regress_stim_base 5
-regress_stim_file 6 $stimdir/Motion_Pre-moco/${subj}_${sess}_motion.1D'[3]' -regress_stim_label 6 'x' -regress_stim_base 6 \
-regress_stim_file 7 $stimdir/Motion_Pre-moco/${subj}_${sess}_motion.1D'[4]' -regress_stim_label 7 'y' -regress_stim_base 7 \
-regress_stim_file 8 $stimdir/Motion_Pre-moco/${subj}_${sess}_motion.1D'[5]' -regress_stim_label 8 'z' -regress_stim_base 8 \
-regress_bandpass 0.008 0.07 \
-scale_max_val 100
-regress_errts_prefix ${subj}_$(sess)_prepro_bp07 \

end

end


Note: EPI datasets are not all of the same length so if any of these -options dont work please notify me.

Thanks for the ongoing help. Really don't know what I would do without it.

~`Dane



Edited 1 time(s). Last edit at 07/05/2014 09:20PM by d6anders.