I will not say i understand it, but when an MRI aquires an image it is a rich k-space dataset. This data is then reduced to a single number (ie intensity) for each voxel in a MR image. Usually the k-space data is discarded after it is used to predict the voxel's intensity.
