I think the warning message in the second output is a red herring; the
dwi2response tournier algorithm is configured to always operate on image data from a single b-value, and if it is provided with input data containing multiple b-values, it will automatically select the largest b-value present.
The main issue seems to be that
dwi2mask is producing an empty mask. Rather than assessing this within the context of the
dwi2response script, you should instead try running
dwi2mask directly on your DWI data and see what happens; even run that command with the
-debug flag to see if anything unusual pops up.
dwi2mask is nowhere near as robust as we would like it to be, but we have been unable to commit resources to implementing something better. Worst case scenario, you can generate a mask yourself and provide that to
dwi2response using the
-mask option; that way,
dwi2response will not need to call
dwi2mask to generate a brain mask, as you would be providing it with such a mask explicitly.