Unsupervised Learning for magnetization transfer contrast MR fingerprinting (MTC-MRF) and chemical exchange saturation transfer (CEST) MRI

Abstract

MTC and CEST MRI are able to image properties of macromolecule and mobile metabolites which cannot be imaged by conventional proton MRI based on water signal. Saturate the magnetization of target compounds with continuous RF pulse and measure how much of the saturated magnetization has been transferred to the water. Most currently used MTC/CEST imaging protocols depend on the acquisition of qualitative weighted images, limiting the detection sensitivity to quantitative parameters, their exchange rate and concentration. Here, we propose a fast, quantitative 3D MTC and CEST imaging framework based on a combined MR fingerprinting and deep-learning techniques. Quantify the MTC parameters with MR fingerprinting and utilize the quantified parameters and Bloch equation signal model to CEST imaging. Especially, the proposed method is based on unsupervised learning that does not require ground truth data. This is highly efficient in MTC/CEST MRI field where ground truth data is limited. The MTC parameter values estimated by the unsupervised learning method were in excellent agreement with values estimated by the conventional Bloch fitting approach, but dramatically reduced computation time by ~130-fold.