Do you have the original scans from which the .img volume was constructed? There are 2 possibilities that I see:
(a) the .img file was improperly constructed from the original data
(b) the wraparound artifact is present in the original data, which can happen due to improper setup of the MRI scan parameters
I would tend to guess the latter, or something like it, since the bottom part of the head (appearing in the top part of the images) when displaced to the bottom would seem to line up pretty nicely -- which is exactly the image artifact that would arise from bad image acquisition or bad image reconstruction.
If the original images are bad, then there is very little that can be done to fix things. In this case, "very little" means "nothing", unless the original k-space data is available.
The stretch-out of the lower part of the head is due to the MRI gradients being nonlinear in that region, and is a separate artifact.