Estimation Of Cardiac Motion Of Cine-MRI Sequences Based On Monogenic Signal

Magnetic resonance imaging(MRI)can image the soft tissue of human body without invasion,and has become an important means of clinical diagnosis of heart diseases.Cine-MRI can dynamically cover the process of myocardial motion and it performs better than tagged-MRI in some aspects,so the research on cardiac motion estimation of cine-MRI is of great value.However,the motion estimation based on cine-MRI is difficult because of the similarity of the gray information and the sparse features.There are still no high precision,fast speed and robust cardiac motion estimation methods of cine-MRI now.Optical flow algorithm is a classical method of motion estimation.It has been applied in the field of medical image processing.To obtain accurate results of cardiac motion estimation based on cine-MRI,more image features description methods need to be introduced into the optical flow.Monogenic signal contains the local amplitude,local phase,and local orientation of an image independently.It can provide more available information for myocardial motion estimation based on cine-MRI and obtain more accurate,fast and robust estimation results.In this paper,we combine monogenic signal and optical flow algorithm to propose an improved algorithm for motion estimation of cine-MRI.The specific research work is summarized as follows:Considering for the characteristics of cine-MRI,we use the monogenic signal features to construct the 3D matrix and introduce them into the optical flow algorithm of the correlation transform.Three local monogenic features of cine-MRI are independent of each other,and the features can be considered as separating the original image into three uncorrelated images,and each image contains local features of pixel location information.These three features are introduced to the optical flow algorithm by the correlation transform method,these features will be the constraints condition in the optical flow and make the uncompleted equation of optical flow became complete,then better estimation results will be obtained.The results based on simulated data show that the proposed algorithm can significantly improve the motion estimation accuracy compared with the classical optical flow algorithm.Experiments based on clinical data show that the algorithm can highlight the areas of abnormal myocardial motion,which are consistent with the location of lesions diagnosed by professional physicians.In order to reduce the calculation time of myocardial motion estimation in the cine-MRI,the Alessandrini algorithm is introduced in this paper.The algorithm is based on the monogenic phase matching and makes an improvement for the estimation of myocardial motion in a tagged-MRI.Compared with the traditional optical flow algorithm,the algorithm not only improves the accuracy,but also calculates obviously faster.However,due to the sparse texture features of cine-MRI,this method is not suitable for motion estimation of cine-MRI.Therefore,the fractional differential method is introduced to enhance the texture of cine-MRI and increase the texture feature of cine-MRI.The fractional differential can effectively enhance the texture information of the image while retaining the information of the image structure,and can get more monogenic phase information for the optical flow algorithm.The simulated data experiments show the algorithm can improve the accuracy of motion estimation and also use short calculating time.Through the validation of clinical data of cine-MRI,the algorithm can accurately diagnose the myocardial infarction area.In order to adapt to the noisy cine-MRI,in this paper,we decompose the original cine-MRI into two parts: texture image and structural image.According to the characteristics of texture image features and image structure,we use monogenic curvature tensor optical flow algorithm for the motion estimation of texture image and optical flow algorithm based on the gray information for the motion estimation of structural image,then we average the two displacements to get the final results.The main idea of this method is that the image is divided in to different components,so that they can be processed separately according to the characteristics of different components.The accuracy of calculated results has been significantly improved,and the robustness is enhanced.The results obtained in the clin ical data can be used to accurately identify the diseased areas.The purpose of this study is to apply the optical flow and monogenic signal for cardiac motion estimation based on cine-MRI and apply the results of estimation for the diagnosis of myocardial motor function related diseases.However,while the results of the motion estimation of cine-MRI are obtained,the diagnostic criteria need to be proved.By comparing the results of cine-MRI and tagged-MRI,we summarize the method for medical diagnosis based on the motion estimation results from the clinical cine-MRI,and the method is verified in several clinical data groups.