A virtual frequency selective inversion (VFSI) method receives at least one MR image representative dataset and an associated phase reference dataset, and classifies anatomical material into a first component representing anatomical material having a first range of resonance frequencies associated with a first range of phase differences between the MR image representative dataset and the associated phase reference image dataset, and a second component representing anatomical material having a second range of resonance frequencies associated with a second range of phase differences between the MR image representative dataset and the associated phase reference image dataset. The method assigns different visual attributes to first and second components derived using phase differences between the MR image representative dataset and the reference image dataset and displays an image.