MR imaging near metal: The POP algorithm

An increasing number of people are receiving joint implants and other metallic prostheses. In many cases follow-up care is required, with consideration given to implant replacement. Magnetic resonance (MR) imaging is well suited to observe the changes to tissue that might indicate complications, yet the images obtained by MR are badly affected by the presence of metal. We are investigating the use of the three-point Dixon technique [1] to reliably separate fat and water near metal implants, but this requires a robust phase estimation method. Current methods fail near the tissue-metal boundary where the magnetic field varies rapidly over a small distance. The POP (phase recovery by onion peeling) algorithm for a 2D image estimates phase around a series of closed paths surrounding the implant. Acceptable results have been achieved by using a Fourier basis for the phase along each path, with the phase estimation problem reduced to refining the Fourier coefficients by minimizing an objective function. Results are shown for two sets of phantom data. POP is currently being extended to 3-dimensional imaging and the algorithm for 3D POP is presented.