Research On Phase Unwrapping In Water And Fat Separation Based Phase Image Quality

Magnetic Resonance Imaging (MRI) is a kind of imaging technology, which based on Nuclear Magnetic Resonance (NMR) in the nucleus. As an electromagnetic signal, the magnetic resonance signal is detected by plus a gradient magnetic field. The signal which including the object structure, physical and chemical characteristics information will be converted into images by the computer system. Compared with other medical imaging techniques, such as X-ray imaging, CT and ultrasound imaging, magnetic resonance imaging technology has many advantages, such as no damage to human body, also do not produce ionizing radiation, and has many contrast imaging, and can choose different imaging parameters and different slices according to the different needs. Therefore, MRI has been widely applied in clinics and maybe the most promising non-invasive diagnostic tool in the diagnosis of human diseases.Magnetic resonance imaging is usually imaging of hydrogen protons within human tissue and the hydrogen protons mainly in water molecules and fat molecules. The fat signal displayed brighter compared with other signals in the image with its adipose tissue of high proton density, short longitudinal relaxation time and long transverse relaxation time. The image of water is usually our interest. Fat signal of high brightness will reduce the contrast of the image, Conceal potential lesions and produce motion artifacts; because proton precession frequency has slight difference in different molecules (that is chemical shift effect), tissue rich in fat will produce chemical shift artifact. On the other hand, in many clinical applications, the fat signal is also a very important diagnostic indicator, tissue contains fat or not, or the proportion of fat in tissue is very important to the judgment and diagnosis of lesions, such as diagnostic bone marrow disease, liver disease and heart disease. In order to meet clinical needs, imaging which suppress fat signal or separating of water and fat is very meaningful. This is because:the fat signal should be suppressed to enhance the contrast of the image signal of water, so to reduce the motion artifacts and Chemical shift artifact caused by fat signal and can be conductive to observe the fine structure of the tissue. At the same time, the image which its fat signals were separated can qualitatively analysis the fat-related diseases. In MIR, the hydrogen protons in water and fat proton have a difference in resonance frequency because of its different molecular environment. That is called the chemical shift effects (Chemical Shift Effect). Using this feature, water signal may be separated from the fat signal.Currently, three-point Dixon method is one of most effective methods for water and fat separation technology. Compared with other traditional methods, such as frequency Fat saturated (Fat-Sat), short TI inversion recovery (STIR), three-point Dixon method has many advantages. Dixon method does not need to interference the pulse on the longitudinal magnetization by applying a pre pulse, so it will not affect the longitudinal magnetization of the organization. The shielding effect of proton by electron cloud around is different between in the fat molecular and water molecular, due to the molecular structure is different. And in the same field, the fat proton precession frequency to be slightly lowers than the water. According to the definition of chemical shift, the frequency difference of between fat proton and water proton is about3.5PPM, that147Hz/T. Dixon method utilized two acquisition where in the phase shift between water and fat signals was adjusted to be0and π (in phase and out of phase), respectively. By adding and subtracting the two images obtained from these data, water and fat image were obtained. When field in-homogeneities are present, the accuracy of the decomposition is diminished since the algorithm can not separate phase shifts caused by BO errors from those deriving from the fat/water chemical shift. Now we need to collect another image to calculate the magnetic field in-homogeneity. However, phase of magnetic resonance system are limited to the range (-π,π], there is phase wrapping and it will be lead to fat and water swap. In conclusion, phase unwrapping is most important in three point Dixon method. Phase unwrapping is in general only considered for an ensemble of pixel for which the true underlying phase is assumed to be spatially continuous or smooth. The method of phase unwrapping is very much, which is divided into three categories. The first category of methods for phase unwrapping is the path-following methods, all the poles need to be first identified either explicitly or implicitly and then properly handled during the path integration. Another general category of methods for phase unwrapping is the minimum-norm methods. The general idea behind the minimum-Norm method is to unwrap the phase by minimizing some cost function. The third general category of methods for phase unwrapping is the maximum flow or minimum cut methods. Phase unwrapping problem into solving a network maximum flow or minimum cut problem, by introducing graph theory maximum flow model. But there’s no way to effectively solve the problem of phase wrapping.We propose a phase unwrapping method based on phase image quality. Phase image are generally considered smooth although affection with magnetic field in-homogeneities. The actual phase image obtained is unsmooth, however. If the quality of the image is related to the changes of the phase gradient adjacent pixels, the points of phase disturbance and interference is poor quality image, and relatively continuous phase image pixels high quality image. In high quality image, phase unwrapping between the regions, and then in the poor quality, phase unwrapping between the pixels and previously unwrapping phase as a guide. The whole process has four steps. First step, in order to reduce the effecting of phase noisy, the phase data is filter by a non local means filter in complex fixed. Second, the phase map is divided into two parts, high quality regions and poor quality regions, with a gradient threshold. At the same time, the high quality regions include many independent regions, which there is no phase wrapping with all the pixels in each regions. That is to say, phase wrapping may be present only between each region. Third, we will increase or decrease integer multiples In each region to make all the regions most closing to a quadric surface smoothly. The mathematical expressions is Φ(t)=Φw(t)+2kπ, and k:=0,±1,±2,…. Last, we phase unwrapping between the pixels and previously unwrapping phase as a guide in the poor quality regions.In this work, with the method of the Fourier transform translation properties, we are Translate the k space data to reduce the number of the stripes and reduces the complexity of the phase unwrapping. Then, image data is filtered by the methods of the non local means filter in the complex fixed, which equal to magnitude and phase filter, respectively. It is both good retained the image edge and the outline of phase wrapping, and go in addition to the influence of noise on image gradient after the filter. Phase unwrapping with high quality image regions can be ensuring the phase information of the same area remained unchanged, and avoiding the inconsistency between adjacent pixels.Phantom and clinical data are used to verify our experimental methods. Our laboratory equipment is XGY-OPER-0.35T permanent magnet-type MRI equipment which is provided by Ningbo XGY magnetic material Co., Ltd, with its SE-Dixon scan sequence, knee coils, scanning parameters TR/TE=580/28ms FOV=240×240, FA=90°, Matrix=256×256, NEX=1, Thickness=8.0mm. Phantom is composed of10centrifuge tubes filled with various proportion of pure water and soybean, in which two centrifuge tubes for comparison:one centrifuge tube is full of water, the other is full of soybean oil. We add an appropriate amount of surfactant Titanium methyl sulfate (SDK,15mmol/L) in solution to make the water and oil uniformly miscible. We also add a small amount of paramagnetic copper(II) sulfate pentahydrate (CuSO4·5H20,0.5g/L) to shorten the relaxation time, add about40g/L of agar as a coagulant. Our clinical data is obtained from14volunteers with a total of340data images by scanning the transverse and sagittal plane of their right leg knee in which the acquisition layer is10, the right ankle sagittal acquisition layer is5. By analyzing the phantom data results, the image brightness becomes brighter along with the concentrations of oil rising. The control group which is full of water, the Signal strength becomes0in the fat image; the tube full of the oil which the Signal strength becomes0in the water image, indicating the water and fat are completely separated. The phase unwrapping is well done, and the separation of water image and fat image as good as wheat in the clinic data. In water image, the image contrast between the organizations is more apparent without fat signal. And distinguish good muscle tissue and bone. Articular cartilage and surrounding connective tissue images is more apparent. There is more convenient to the doctor’s clinical diagnosis. We can be according to the brightness of the image to estimate fat distribution and content within the organization from fat images.