Phase Unwrapping for Magnetic Resonance Thermometry

摘要

Magnetic resonance thermometry enables temperature measurement in a non-invasive way. A representation of the temperature is given by the phase shift of the MR signal. Unfortunately, only "wrapped phase" is measured directly, yielding values between -π and +π. However only the true phase(" unwrapped phase") is proportional to temperature. Phase unwrapping is essentially the process of recovering true phase from its wrapped representation. The unwrapping boils down to adding a multiple of 2π (the wrapped count) to each region enclosed by "fringe lines" marking a jump. In practical phase images, however, noise can cause such lines to merge, break them into parts, or create isolated segments. Locations where fringes merge or segments end are called residues or discontinuity sources. The problem of phase unwrapping is often solved by minimizing the number of discontinuities (phase jumps) over all choices of wrap count. Unfortunately, the minimum discontinuity solution is not always appropriate for phase data that contains noise. To solve this problem " quality map" can be used to weight the discontinuities. Any further spurious discontinuities may be reduced by low pass filtering the complex image. In this paper we will investigate the effect of quality maps derived from MRI magnitude images. Additionally, we will look into noise filtering of the complex MRI images.