Two-point Dixon fat-water separation: improving reliability and accuracy in phase correction algorithms.

PURPOSE: To propose an advanced phase-correction region-growing algorithm for two-point fat-water separation suitable for parotid assessment, and to evaluate the general performance of phase-correction algorithms.
MATERIALS AND METHODS: Two region-growing algorithms were evaluated in test objects and in head images: the original phase-correction algorithm (OPC) and the advanced phase-correction algorithm with voxel size manipulation (VSM) which includes: 1) starting the region-growing process from images of lower resolution and subsequently stepping toward the original matrix size, and 2) limiting the use of low-pass filters to fat-water interfaces with partial volume effects
RESULTS: Fundamental problems relate to biological tissue spectrum being poorly approximated by two discrete peaks for fat and water. The VSM algorithm was shown to be less noise-sensitive, faster, and to produce a better approximation for the field inhomogeneity map. In head images (6 volunteers, 10 slices each) 43 errors were found with the OPC algorithm and only 6 errors with the VSM algorithm. Only the OPC algorithm produced errors surrounding the parotids (10 errors).
CONCLUSION: The VSM algorithm provides a more accurate and less noise-sensitive fat-water separation. This highly significant performance improvement allows the application of phase-correction algorithms to a wider range of clinical applications.