ROI analysis details

HI List experts,

I was wondering whether I can do the ROI analysis based on the output from 3dDeconvolve (fixed-shape regression, for instance, BLOCK(d,p)). This is related to the step of “3dROIstats”. Any good pointers?

Thank you very much,
Juan

Juan, it’s hard to offer suggestions because it is not clear as to what you’d like to achieve through ROI analysis at the individual level.

Gang,
Sorry I did not say it clear. My question is, when we do the 3dDeconvolve, we can use fixed-shape regression, such as BLOCK(d,p) method for each subject. Can I use its output for the ROI analysis. Let us have two 3dDeconvolve as examples below.

#1 shape fixed
3dDeconvolve -input rall_vr+orig
-concat ‘1D: 0 108 216 324’
-num_stimts 4
-stim_times 1 ‘1D: 17.5 | 185.0 227.5 | 60.0 142.5 | 227.5’ ‘BLOCK(20,1)’ -stim_label 1 Actions \
-stim_times 2 ‘1D: 100.0 | 17.5 | 185.0 227.5 | 17.5 100.0’ ‘BLOCK(20,1)’ -stim_label 2 Tools
-stim_times 3 ‘1D: 60.0 227.5 | 60.0 | 17.5 | 142.5 185.0’ ‘BLOCK(20,1)’ -stim_label 3 HighC
-stim_times 4 ‘1D: 142.5 185.0 | 100.0 142.5 | 100.0 | 60.0’ ‘BLOCK(20,1)’ -stim_label 4 LowC \
-gltsym ‘SYM: Actions -Tools’ -glt_label 1 AvsT
-gltsym ‘SYM: HighC -LowC’ -glt_label 2 HvsL
-gltsym ‘SYM: Actions +Tools -HighC -LowC’ -glt_label 3 ATvsHL
-fout -tout
-bucket func_rall_Shape_fixed
-cbucket coef_rall_Shape_fixed_betas

Mostly, we also can do it, using Non-fixed shape regression, such as CSPLIN for each subjects as following.

#1 shape NOTfixed
3dDeconvolve -input rall_vr+orig
-concat ‘1D: 0 108 216 324’
-num_stimts 4
-stim_times 1 ‘1D: 17.5 | 185.0 227.5 | 60.0 142.5 | 227.5’ ‘CSPLIN(0,15.6,7)’ -stim_label 1 Actions \
-stim_times 2 ‘1D: 100.0 | 17.5 | 185.0 227.5 | 17.5 100.0’ ‘CSPLIN(0,15.6,7)’ -stim_label 2 Tools
-stim_times 3 ‘1D: 60.0 227.5 | 60.0 | 17.5 | 142.5 185.0’ ‘CSPLIN(0,15.6,7)’ -stim_label 3 HighC
-stim_times 4 ‘1D: 142.5 185.0 | 100.0 142.5 | 100.0 | 60.0’ ‘CSPLIN(0,15.6,7)’ -stim_label 4 LowC \
-gltsym ‘SYM: Actions[2…3] -Tools[2…3]’ -glt_label 1 AvsT
-gltsym ‘SYM: HighC[2…3] -LowC[2…3]’ -glt_label 2 HvsL
-gltsym ‘SYM: Actions[2…3] + Tools[2…3] -HighC[2…3] -LowC[2…3]’ -glt_label 3 ATvsHL
-fout -tout
-bucket func_rall_Shape_NOTfixed
-cbucket coef_rall_Shape_NOTfixed_betas

In the past, When I do the ROI analysis, my input is usually from the “func_rall_Shape_NOTfixed”. Now I find the results from certain time points (e.g, 2.5 s since the onset of the stimulus) is not good, I hope to see the ROI results from the 3dDeconvlve of shape non-fixed one, i.e., func_rall_Shape_fixed.

Is this possible? Any comments?

With CSPLIN(0,15.6,7) as your basis function, you could plot out the response curve or shape at the region or ROI level. On the other hand, with BLOCK(20,1), a ROI analysis would only allow you to see the estimated response amplitude since the response shape is fixed as the gamma variate convolved with the duration of 20s.

If CSPLIN(0,15.6,7) is your real basis function for a block lasting for 20s, that could be troubling because a 20s block presumably would have response lasting for 30s or more.