Accelerated model-based proton resonance frequency shift temperature mapping using echo-based GRAPPA reconstruction

Abstract

PurposeTo develop an acceleration method for MR temperature estimation using model-based proton resonance frequency (PRF) shift method. Materials and MethodsImages of 16 different echo times (TE) were acquired in one RF excitation using a multi-echo gradient-recalled echo (GRE) sequence. Fully sampled k-space data were retrospectively under-sampled at a net reduction factor between two and three using the proposed under-sampling strategy. K-spaces of three different TEs were combined together to perform the proposed reconstruction method called Echo-based GRAPPA. Ex vivo goose liver cooling experiment and in vivo breast imaging experiment were performed to investigate the accuracy of Echo-based GRAPPA. Conventional GRAPPA reconstruction was implemented for comparison using the same sampling pattern. ResultsThe goose liver imaging experiment shows that the reconstruction-induced temperature RMSE of a selected region of interest (ROI) is less than 1.4°C for Echo-based GRAPPA at a net reduction factor of 2.3. The breast imaging experiment shows that the mean temperature error of water–fat mixed ROIs is 2.3°C at a net reduction factor of 2.7. Conventional GRAPPA shows larger temperature RMSE than Echo-based GRAPPA. ConclusionThe proposed method can accelerate the MR temperature estimation using model-based PRF at a net reduction factor between two and three with a reconstruction-induced temperature error less than 3°C in water–fat mixed ROIs.