The Research Of MRI Preprocessing Based On Image Decomposition

In recent years,the development of medical imaging technology has provided a reliable basis for the diagnosis and treatment of diseases.The magnetic resonance imaging(MRI)has become one of the important methods for the clinical examination of liver and breast tissues,because of its high soft tissue resolution,spatial resolution,and no radiation damage.Due to the intensity inhomogeneity of the radio frequency field and the patient's respiratory motion during imaging,the MRI will respectively generate the bias field and the displacements between different phases,which seriously affects the subsequent image analysis and processing.Therefore,to remove the above mentioned bad influences,the image preprocessing work is particularly necessary.Image preprocessing is the basic work of image processing,and its results will have a great impact on the subsequent analysis and processing of the image.This paper mainly focuses on the image preprocessing,which contains the correction of bias field in liver MRI and the registration of displacement deformation between different phases of the breast dynamic contrast-enhanced MRI.Generally,the bias field of the liver MRI is regarded as a multiplicative field.To achieve the image with the bias field correction,the image decomposing technique can be applied to separate the bias field form the liver MRI.The registration of the breast dynamic contrast-enhanced MRI is a difficult problem in the field of medical image registration.The intensity variation between different phases,which is caused by the diffusion of contrast agents,makes the registration problem is not easily resolved.In this paper,the image decomposition technique is used to decompose the breast dynamic contrast-enhanced MRI,and the additive components,the normal tissues and the abnormal tissues,can be separated.Then,the displacement deformation fields are extracted using the normal tissue image with gray consistency and is applied to realize accurate alignment of the breast dynamic enhanced MRI.The main contents of this paper are as follows:(1)The bias field correction of liver MRI based on the entropy minimization and edge-preserving filtering.To correct the multiplicative bias field of liver MRI,a bias field correction algorithm based on the entropy minimization and edge-preserving filter is proposed in this paper.Firstly,the background image noise is removed by the template image.Then the entropy minimization algorithm is used to decompose the image into two parts: the bias field image and the image without bias field.At the same time,the edge-preserving filter algorithm is used to compensate the detail information obscured during the entropy minimization process.And finally the liver MRI images with the bias field correction are obtained.This paper based on the bias field correction experiments of the clinical liver MRI.By comparing with the existing algorithms,the gray and segmentation results and other multi-parameters confirm that the proposed algorithm achieves an ideal correction effect.(2)Non-rigid registration algorithm based on TV decomposition model with space-time dual constraints.For the registration of breast dynamic contrast-enhanced MRI images,this paper proposes a non-rigid registration algorithm based on TV decomposition model with space-time dual constraints.This algorithm decomposes the breast dynamic contrast-enhanced MRI images by the TV model with time-space dual constraints into the abnormal tissues and the normal tissues.The abnormal tissue images contain the gray changes and the normal tissue images have gray consistency and contain displacement deformation.Then the displacement deformation fields are extracted using normal tissue images and are applied to align the original image.Finally the better registration results of breast dynamic contrast-enhanced MRI images are obtained.This paper verifies the performance of the algorithm by using the clinical images and comparing with the existing algorithms.