Additional processing for phase unwrapping of magnetic resonance thermometry imaging

Abstract

Magnetic Resonance Thermometry imaging is a non-invasive method for temperature monitoring in hyperthermia treatment. The temperature can be determined from the phase shift in a gradient-echo sequence. Due to large temperature variations, the phase shift may exceed the (-π ,π) radians interval. The phase value beyond this interval will be wrapped. Unfortunately, the temperature is only proportional to the absolute phase change. Therefore, phase unwrapping (PU) is required to recover the absolute phase from the wrapped representation. While the phase may contain spurious discontinuities, the algorithm must distinguish them from true phase discontinuities. We propose additional processing to support PU in order to improve the algorithm for recovery of the best estimation of absolute phase. The Minimum Weight Discontinuity (MWD) algorithm was used for PU. The steps to be taken on additional processing consist of applying a Gaussian filter to the raw complex MRI images, deriving the weights of a quality map, and segmenting unreliable regions using the magnitude image. The raw wrapped phase images, acquired from a phantom and from a porcine liver (acquired under laser irradiation), were used to test the effect of additional processing. The effect was compared with the conventional approach (i.e. mere unwrapping with the MWD algorithm).