Highly Accelerated Temperature Mapping Using Nonlocal Regularized Parallel Imaging

Sheng Fang¹, Xinyi Pan¹, Kui Ying¹

Model-based MR thermometry method based on the proton resonance frequency shift (PRFS) can effectively improve the temperature estimate accuracy of conventional phase different method. However, its temporal resolution need be improved for real-time temperature monitoring. To solve the problem, we applied highly accelerated PI to temperature mapping and used nonlocal regularization that extracts the prior from the acquired data themselves to stabilize the reconstruction. The method was demonstrated using whipping cream phantom. The results show that the nonlocal regularization can effectively increase the temporal resolution of PRFS while avoiding the introducing the bias to quantification, due to its data-driven property.