This branch was meant to address #7, but in the process of doing that I realized that the parameters generated by the motion_correction workflow were totally screwed up by the use of a nonstandard prefix for the transform files. I ended up changing the model to use the single-timepoint registration (like @lukebrez had originally done), and then to apply the transforms for each timepoint, so that I can then delete the transform files to prevent them from filling up the /tmp directory on the compute nodes. It has been tested on the test dataset and appears to work properly.
This branch was meant to address #7, but in the process of doing that I realized that the parameters generated by the motion_correction workflow were totally screwed up by the use of a nonstandard prefix for the transform files. I ended up changing the model to use the single-timepoint registration (like @lukebrez had originally done), and then to apply the transforms for each timepoint, so that I can then delete the transform files to prevent them from filling up the /tmp directory on the compute nodes. It has been tested on the test dataset and appears to work properly.