For 3D acquisitions, my software may create additional images with the 'o' and 'co' prefix. The raw data is saved without a prefix. The 'o' image is rotated to the nearest orthogonal plane as the NIfTI. The raw and 'o' image are identical for tools that fully support the NifTI format (e.g. SPM12), but the 'o' image might appear in a more canonical orientation if you view it in software that shows how the data is saved to disk (e.g. FSLview). The 'co' image attempts to crop excess neck signal that can disrupt coregistration and normalization. I would visually inspect these images, but the 'co' images are often the ideal choice for normalization to stnadard space or coregistration with images that show less neck signal (e.g. fMRI, DTI).