The Research On Water-Fat Separation Which Based Interval Sampling And Region-growing

Magnetic resonance imaging(MRI),as one of the major medical imaging technologies,it provides lots of information and quantitative information for clinical diagnosis and treatment.Due to the long T2 and short T1 of fat,the signal from fat is hyper-intense in both T1-weighted and T2-weighted images.However,the hyper-intense fat signal will affect the image contrast,even conceal some potential diseases,and affect the clinical diagnosis.Therefore,the fat suppression has a great significance in MRI.Dixon methods not only can get fat suppression image,but also obtain fat image.The two-point Dixon imaging is often desirable at clinical because of its high scan efficiency and flexibility.The two-point Dixon method which based on region growing algorithm,is easy to obtain error result of water-fat separation.that is because the region growing algorithm prone to propagate and accumulate error if the image has low signal to noise ratio,artifacts and motion.In this paper,a new two-point Dixon method that based on interval sampling and region growing algorithm is proposed.The experiments verify that proposed method can decrease error propagation and accumulation,and can obtain correctly results of water-fat separation.The main steps of the proposed method is:Firstly,the four pairs field vector map will be obtained through downsampling the field vector map,the purpose of this manipulation is reduce the effect of phase which with noise and rapid change;Then,the four sub-field vector map phasor maps will be acquired from the four pairs field vector maps by using region growing algorithm,and the four sub-field vector will be correction by an smooth operation,which will make the phase of sub-field vector more accurate;Finally,the four sub-field vector will be combined together,and the final field vector map is obtained.The innovation of this method is:firstly,the proposed method will get four sub-images in the same time through interval sampling,which will provide more image information and the four sub-images can be mutually constrained in the processing;Secondly,the proposed method use the interval sampling to distribute the upheaval phase into four sub-images,and this manipulation decrease the probability of four sub-images have same upheaval phase in the same region;Thirdly,by combining interval sampling and region growth algorithm,the influence of the error propagation and accumulation caused by the region growing algorithm on the final vector map is decreased,the final result will be more robust and reliable;Finally,the four sub-field vector map can be parallel computing,which will reduce the time of calculation.In this paper,the algorithm is verified by simulation and in-vivo experiments.The results of the first simulation phase data show that the proposed method has less error than the original region growing algorithm in the presence of multiple phase changes.The results of noise simulation experiment show that the proposed method is more robust to the data which with noise.The results of the in-vivo water-fat separation experiments show that proposed method is more stable and accurate.The calculation time of the proposed method is more less than the two-point Dixon method which based on region growing,and this advantage will be more obvious if the matrix size is large.