SU-E-J-53: A Phantom Validation Study of a 3D Background Phase Model for MR Thermometry

The proton resonance frequency (PRF) shift method is the most widely accepted method for magnetic resonance thermal imaging to provide real-time treatment monitoring of thermal therapies. However, the PRF shift technique involves the subtraction of a reference phase map, which causes the technique to be easily perturbed by tissue motion and other background contaminations. In this study, a three-dimensional background phase is estimated in order to create a phase reference for each time point.

A magnetic resonance spectroscopy (MRS) sphere was scanned within a 3T MRI scanner employing a 3D fast SPGR sequence. Real and imaginary images were acquired to obtain phase images as the control. The ability to predict the background phase was investigated by systematically removing phase information from the control data set. Data was initially removed from a spherical region of interest (ROI) to simulate a region where ablativeheating would take place. In a second case, the same spherical ROI was removed as well as every other slice to further reduce the amount of existing data. A 3D finite element model was implemented to solve the Dirichlet problem given a measured phase on the boundary of the simulated available data.

Line profiles taken through the phantom indicate phase estimates to compare well with actual phase measurements. The phase estimation still shows good agreement when reducing the amount of data to every other slice.

The 3D multi-slice temperature estimate potentially provides a robust technique that is not as susceptible to through-plane or in-plane motion-induced temperature artifacts as compared to thecurrent PRF shift method. The research in this paper was supported in part through 1R21EB010196-01.