An automated iterative algorithm for water and fat decomposition in three-point Dixon magnetic resonance imaging

An iterative, outlier exclusion, second-order surface fitting algorithm has been developed to solve the well-known phase wraparound problem associated with in vivo applications of the three-point Dixon magnetic resonance imag ing method. The technique was optimized for speed by reducing the problem t o a pair of planar fits. The spatial misalignment between water and fat com ponents due to the chemical shift was handled on a subpixel level by invoki ng the shift theorem of Fourier transformation. From the chemical shift cor rected water and fat images, high quality recombined MR images were generat ed. The algorithm was validated in both phantom and patient studies. In viv o breast images and pelvic images are provided as a demonstration of the me thod. (C) 1999 American Association of Physicists in Medicine. [S0094-2405( 99)03911-5].