The results that have been presented here are highly dependant on the accuracy of phase unwrapping procedure. For all images tested, the unwrapping process in the brain region was almost always successful, but problems were encountered for in thin region defined by the skin. So, more work is required to perform perfect unwrapping of the phase over skin-regions.
Another problem experienced was movement during acquisition, causing the three-point Dixon technique to fail whenever a geometrical inconsistency was found between the three images. While we had to acquire the three images in three separate acquisitions, an appropriately interleaved acquisition (unavailable to us) could alleviate this problem. In our case, a rather long time interval (in the order of 2-3 minutes) in between the three scans of the three-point Dixon technique, made the movement difficult to avoid.
Because of these factors, it was difficult to perform quantitative analysis of the field offset caused by susceptibility effects at the air-skin interface. We feel however that a better unwrapping and a suitable acquisition sequence will make it possible.