Rearch On Magnetic Resonance Fat Quantification And Its Application

Non-alcoholic fatty liver disease（NAFLD）and its progressive subtype non-alcoholic steatohepatitis（NASH）have become the major cause of chronic liver disease worldwide over the past three decays.Non-invasively detection of NAFLD and NASH is an urgent.The MRI proton density fat quantification（PDFF）could accurately reflect the degree of histologic steatosis grade in NAFLD,but the robustness of fat-water separation might become challenging.Multi-parametric magnetic resonance imaging（MRI）provides a promising imaging tool for NASH.However,challenges still remain when applied in human abdomen.The presence of fat could lead to a biased T₁ quantification;breath-hold is required in abdominal scans to reduce respiratory artifacts,and the large imaging volume within a single breath-hold requires high acquisition efficiency.To overcome the difficulties in existing methods,the following works have been carried out.Firstly,an algorithm named by“Transition REgion Extraction”（TREE）was proposed in this work to achieve robust fat-water separation.The robustness of fat-water separation algorithms may become challenging in applications with varying inhomogeneity of main magnetic field B₀ and low signal-to-noise ratio（SNR）due to model mismatch.To overcome these difficulties and improve the robustness in region growing methods,modification has been made to traditional region growing methods,and TREE algorithm was proposed in this work to achieve robust fat-water separation.Experiments show that the proposed method could accurately quantify the fat fraction with a bias of only 0.14%.Compared with traditional methods,the TREE algorithm was more robust in the applications with spatially varying B₀ and low SNR.Secondly,this work optimized the sequence of abdominal multi-parametric imaging.The acquisition efficiency of existing multi-parameter quantification methods was relatively low,which leads to a prolonged scanning time or limited imaging volume.In this work,the quantification sequences have been optimized.The B₁+inhomogeneity has been corrected and the effects of fat signal on quantification has been removed with combination of fat-water separation algorithms.Simultaneous T₁/PDFF/R₂^*quantification in the whole liver within a single breath-hold has been achieved.Besides,exploiting the difference of T₁ relaxation time of fat and water signal,a possible way for fat-water ambiguity solving has been proposed.This work could accurately quantify T₁/PDFF/R₂^*of the fat-water phantom.Compared with the golden standard,PDFF bias was only 0.73%,T₁ bias was only 4.3%,R₂^*bias was only 4.07 s^-1.On the basis of multi-parametric imaging methods,animal experiments were designed and implemented to explore the feasibility of applying proposed method in early detection of NASH.Based on pathological results,the proposed MRI parametric imaging methods could be used to stage the inflammation in NASH model,the area under receiver operation curve for distinguishing inflammation grade 0-1 and≥2 could reach 0.79.The establishment of diagnostic model provided a sound basis for the following clinical applications.In summary,the fat-water separation methods have been modified and the robustness has been improved thereby.On the basis of fat-water separation,the abdominal multi-parametric quantification sequences have been optimized,and multi-parameter quantification in whole liver within a single breath-hold has been achieved.Finally,in the animal models,the diagnostic value of multi-parametric quantification has been validated,which laid the foundation for further clinical trials.