Jason-
  The latest versions of dcm2niix are more sophisticated. It will rotate 3D acquisitions to the nearest orthogonal to a NIfTI identity matrix. If you request the crop ('-x y') than you should get both the original as well as a cropped scan.

In addition, tools like BET have become better - for T1 scans I have been impressed by the "Robust" (-R) option.

Finally, for my own data I use SPM to compute how to align the T1 scan to standard space and then impose a bounding box. Crucially, like dcm2niix this cropping is done losslessly for the surviving voxels: the data is not resampled. 
  https://github.com/rordenlab/spmScripts/...