Fat quantification in liver using chemical shift imaging

Abstract

MRI is becoming increasingly important in the characterization of NAFLD, which is one of the most relevant di use liver diseases with increasing prevalence in the developed countries. Currently, biopsy is the gold standard in the early identi cation and staging of NAFLD, which is crucial to evaluate the risk of hepatocellular carcinoma development. However, the later is associated with sampling errors and with the development of several post-surgical complications. In this work, we address the problem of quantifying the fat accumulation in the liver using a non-invasive procedure based on chemical shift imaging and two signal models, the magnitude and complex models. In particular, we address the in uence of noise, relaxation e ects and choice of echo times in fat fraction estimation error. Our approach relied in the development of a new acquisition strategy based on the optimal choice of echo times, which were evaluated in simulation studies and later tested with phantoms. A program to perform in-vivo fat quanti cation was implemented, extending our approach to clinical data. Results indicate that good fat fraction estimation can be achieved by correctly choosing the echo times. However, it is extremely di cult to nd a single echo time combination which optimize a large range of fat fractions. Finally, we also demonstrate that the standard clinical protocol has several aws and that a new acquisition protocol must be developed.