Issues using 3dDeconvolve on sparse-sampling fMRI data

Hello!

I am attempting to use AFNI’s 3Ddeconvolve function to estimate beta weights per voxel per trial for some sparse-sampling fMRI data collected by the speech-motor neuroscience lab I work in. Eventually we would like to perform this step in the context of a Nipype workflow, but I’ve just been playing around with the function in the command line with a single subject and for one condition first.

I’m providing 5 functional nifti files as input, each corresponding to one ‘run’ of 64 trials. Listen.1D is 5-by-24 set of times at which the stimulus occurred for the listen condition, local to each run.
There is an initial scan at the beginning of each run that I am trying to use CENSORTR to censor out as there would not have been any stimulus at that point.

Attached are the error I’m getting and the command I’m running, and I’m wondering if someone could help me determine why I’m receiving these matrix issues and perhaps help me to fix them.

Thanks!

-Nickolas

Nickolas,

The warning message indicates that 4 columns in the design matrix are all 0s. Add option -x1D in your 3dDeconvolve script, and spill out the design matrix. Then use tools such as 1dplot, ExamineXmat and xmat_tool.py to figure out what those zero columns correspond to in your stimulus timing file Listen.1D.