Greetings All,

We've come across an interesting challenge. We're extracting impulse response data from a 4D dataset using 3dDeconvolve and condition onset vectors associated with our paradigm. We've used, on a number of occasions, a script like the one that appears at the end of this message and have come up nothing but aces (i.e. very nice functions in all the right places).

However, the results of running this same process on a new data set have given us trouble. Specifically, the waveforms that result have lots of zeros and a few non-zero values, is in:

voxel(32, 26, 4) = [0 -.53 0 0 0 0 0 0 0 0 0 0 -.36]
voxel(32, 27, 4) = [0 -.53 0 0 0 0 0 0 -.38 0 -.38 0 -.36]
...

(see attached graph pic)

Some information that may be useful in solving this:

1) The zeros are not "close to zero;" they're just null.
2) When a non-zero value occurs, it's always the same value across all the voxels from that sub-brick (a.k.a.: time-point, in this case).
3) *None* of the preceding holds for voxels outside the brain. They look like their normal noisy selves.

This last point, in concert with the fact that we convert our data to percent change before modeling (thereby making within-brain--but not outside-brain-- BOLD close to zero), makes me wonder if there hasn't been a change to some rounding procedure in a recent AFNI rev that doesn't work well with our configuration (Version AFNI_2008_07_18_1710 [[Precompiled binary macosx_10.4_Intel: Oct 21 2008]]).

Any input on this disturbing problem will be very well-received. Please let me know if I can provide additional information of use...

Best,

Paul


3dDeconvolve -jobs 4 -input 4D_data+orig -nfirst 0 -num_stimts 9 \
-polort A \
-stim_file 1 '3dmotion.1D[1]' -stim_label 1 'roll' \
-stim_file 2 '3dmotion.1D[2]' -stim_label 2 'pitch' \
-stim_file 3 '3dmotion.1D[3]' -stim_label 3 'yaw' \
-stim_file 4 '3dmotion.1D[4]' -stim_label 4 'SI' \
-stim_file 5 '3dmotion.1D[5]' -stim_label 5 'LR' \
-stim_file 6 '3dmotion.1D[6]' -stim_label 6 'PA' \
-stim_file 7 vector_condition1.1D.irf -stim_label 7 'Condition1' \
-stim_file 8 vector_condition2.1D.irf -stim_label 8 'Condition2' \
-stim_file 9 vector_condition3.1D.irf -stim_label 9 'Condition3' \
-iresp 7 Condition1 -stim_maxlag 7 14 \
-iresp 8 Condition2 -stim_maxlag 8 14 \
-iresp 9 Condition3 -stim_maxlag 9 14 \
-tout -fout