Single image phase-based MRI fat suppression expectation maximization algorithm

MRI fat suppression techniques are primarily approached by variations of the phase based methods. Multipoint methods dominate the literature, specifically methods that require the acquisition of three images at different echo times. These methods estimate the contribution of fat and water signals by solving a set of equations using echos acquired with the relative phase of fat and water at -π, 0, π. We present a novel approach to fat suppression that requires only a single image. Using an expectation maximization algorithm on a gradient echo image with the relative phase of fat and water at approximately π/2, the pixel composition and inhomogeneity of the magnetic field are estimated. Fat suppressed images are generated from estimated inhomogeneity. Our method does not need custom pulse sequences. Results from mathematical phantoms and volunteer images are reported.