I've been using dcm2niix for awhile to convert dicoms from multiple scanners to niftis for use in our research. Some of those dicoms are ADC or DWI, or contain multiple scans in a single stack which understandably creates a 4D file that I spilt into multiple components. However, I ran across a 376 slice image from a GE scanner that dcm2niix is making into a 4D file and I can't figure out why. All of the slices have the same pixel dimensions and echo times, and the only difference is the positioning data. Dcm2niix splits the image into two 188-slice images, which result in each image showing only half the brain. 

Is this a known problem with a work around? If so, please share. If not, how do I prevent the split. I can create a single 3D image, but I have no way of knowing when I should do that versus create separate niftis out of the components.

Is it possible this comes from a new x2 multi band sequence. The issue is that this format did not exist until recently. See here for details.
  https://github.com/rordenlab/dcm2niix/issues/336
  https://github.com/rordenlab/dcm2niix/is...
If this is correct, you can convert these images using the new developmental release. You could build this on Unix with the command:

git clone --branch development https://github.com/rordenlab/dcm2niix.gi...
cd dcm2niix/console
make
./dcm2niix

If you have Windows, a compiled 'Artifact' of the development branch is created byAppVeyor
  https://ci.appveyor.com/project/neurolabusc/dcm2niix

I did the rebuild on MacOS and the problem was indeed solved. I need to do the build on my Docker containers which might take a bit more work as they start out pretty bare.

Thank you!