Digital Image Enhancement Based On Quantitative MRI

Quantitative magnetic resonance imaging (MRI) can measure a variety of human physiological tissue parameters quantitatively,providing a more objective basis for clinical diagnosis of diseases.Additionally,synthetic MRI can generate contrast weighted images with the governing equations based on measurement of tissue properties (PD,T1,T2) and the synthetic images are similar in appearance to images normally acquired with an MRI scanner but with improved acquisition speed.Two main studies on the related topics were performed in this thesis work:To address the problems that the conventional synthetic MRI images are still inferior in general to real MRI images in contrast and ability to detect diseases,we proposed an enhanced synthetic MRI method,which could greatly improve the quality and contrast of synthetic MRI images and promote better detection power for clinical diseases.Quantitative susceptibility mapping (QSM) can generate a three-dimension susceptibility distribution and evaluate iron content and degree of calcification.Given the problem that we cannot reconstruct high quality QSM maps from data with short acquisition time by using traditional methods,we tested the feasibility of using a 3D U-Net neural network to reconstruct QSM maps with single echo gradient echo (GRE)phase data,low-resolution single echo GRE phase data,and susceptibility weighted imaging (SWI) phase data.The experimental results suggested that the outcomes from reconstruction of single echo GRE phase data by neural network were close to those by traditional method,and the outcomes of reconstruction by neural network from low resolution single echo GRE phase data and SWI phase data were clearly better than those by traditional method.In summary,we improved the image quality of quantitative MRI and its derivatives by improving existing techniques or using neural networks.