Research On Motion Compensation For Intravascular Ultrasound Sequence Based On Nonlocal Means

Abdominal aortic aneurysm is one of the most common aneurysm and a serious threat to human life in vascular disease. Currently, endovascular stenting is the best choice for treatment of abdominal aortic aneurysm. However, since special types of aortic aneurysm did not meet clinical requirements of stenting, conventional stenting for special types has to block the entrance of vessel, but eventually lead to failure of vital organs. Furthermore, the custom stent with branch is expensive and needs a long custom cycle and is difficult to operate in clinic. Therefore, we firstly propose to locate entrances of vessel branches by intravascular ultrasound (IVUS) and then utilize the laser to implement fenestration on stent to solve the problem that important branch arteries are cut off by stent.However, in intravascular ultrasound imaging, due to heart beat and rotation of catheter, IVUS sequence exists motion artifacts. Motion artifacts will directly affect doctors on diameter measurement of vascular branch entrance, which produces negative influence on judgment of entrance of vascular branch to result in difficulty for interventional treatment of laser. Furthermore, motion artifacts will interfere with plaque measurement and3D visualization of vessel. Motion compensation technology for IVUS aims to attenuate motion artifacts in IVUS sequence, which will provide more precise3D structure of vessel and accordingly, conduce to accurate judgment of branch vessels for doctors. Furthermore, motion compensation technique can provide accurate information for plaque and vessel elasticity measurement. Therefore, motion compensation technology is not only an urgent need to solve technical problem of our project, but also has an important clinical significance.Currently, IVUS motion compensation technology is mainly based on gating, kinetics modeling, template matching and optical flow methods. The gating method can only collect one image in one cardiac cycle; kinetics modeling method cannot provide accurate estimation result of rotation angle; template matching method will be affected by change of image grayscale; optical flow method requires the vascular center without displacement. So their results are unsatisfactory in practice. In this thesis, aiming to difficulties of IVUS motion compensation technology and combining with the doctors’clinical demands, we study a novel IVUS motion compensation method based on nonlocal means and concrete research contents are as follows:Firstly, nonlocal means based speckle reduction for IVUS image is studied:based on characteristic of speckle noise on IVUS images, a weight refining based nonlocal means method for despeckling is proposed. Based on the signal-dependent speckle model, a novel similarity weight is derived by Bayesian framework. The weight is iteratively refined in a subspace using principal components analysis (PCA). Weight refining is automatically terminated. This method provides satisfactory results and enhances contrast.Secondly, nonlocal means based coarse motion compensation method for IVUS sequence is researched:according to modeling motion of IVUS catheter, which is divided into rotation due to catheter rotation and translation to image center because of blood pulse, a media-membrane detection based motion compensation method is proposed based on priori knowledge that media-membrane in IVUS image represents ellipse. Proposed method firstly detects edges by nonlocal SUSAN detector, then finds controlling points in region of interest, and finally segments media-membrane using ellipse fitting using controlling points. Initial motion parameters are estimated by fitting parameters of ellipse and final ones are computed by1D nonlocal means on initial parameters and then used for motion compensation. The proposed method executes easily and fast and can be used for various IVUS images.Finally, nonlocal means based precise motion compensation method for IVUS sequence is studied:aiming to the problem of the inaccurate rotation angle estimation of coarse motion compensation method, a nonlocal optical flow based precise motion compensation method is proposed based on the results of coarse motion compensation. The proposed method introduces the similarity of nonlocal means into Lucas-Kanade (LK) optical flow to improve its noise-immune performance. A calculation method for rotation angle vector is also proposed and then the whole process of motion compensation is finished based on the results of coarse motion compensation using precise rotation angle. The proposed method successfully solves the rotation problems of IVUS sequence and it is greatly superior to the traditional method in terms of compensation accuracy.