Reserch Of Intravascular Ultrasound Image Feature Points Matching And Registration

With the improvement of people’s material quality, the incidence and mortality of cardiovascular disease is increasing. Intravascular ultrasound technology is an emerging imaging technology for observe blood vessel, which provides cross-sectional images of the vessel wall. It is widely used in clinical and has become an effective tool in the diagnosis of cardiovascular disease. The research for the feature points matching and images registration based on the intravascular ultrasound can provide sufficient basis for the flexibility analysis and three-dimensional reconstruction of blood vessels, in addition, it is beneficial for observe the therapeutic effect and vascular lesions besides, It also has important significance for guide the interventional operation guidance and monitor the stent implantation surgery.This paper firstly gives a brief description of the research status, purpose and significance of intravascular ultrasound image processing, introduces the definition of image registration, classification and basic steps, and expounds the development process of image registration and related research. The research content and the chapter structure of this chapter is summarized subsequently. This paper also introduces the basic principles of intravascular ultrasound imaging, and adopt a new approach of the intravascular ultrasound image edge detection for complete the ROI segmentation task which lay a good foundation for the later study. The research contents about feature points matching and images registration based on the intravascular ultrasound point image are as follows:(1) In order to find the matching points before and after the deformation more accurate and quicker for obtaining deformation under different blood pressure, thus get better analysis of vessels elasticity. A method of uniform marked points based on region constraint is used in the first time. Firstly, extract uniform marked points of IVUS ROI images, then an points matching method based on the range constraint is proposed for reduce the matching time, which chose the fast cross-correlation function as the similarity measure function and make a seek in the search window, and select the center point of sub-image as the matching points of marked point. The experimental results show that chose the improved normalized as the similarity measure function can achieve better accuracy and less time than other function.(2) Because of the IVUS images have the low contrast characteristic and the SIFT feature have large computation. An improved SIFT feature extraction and matching method based on the IVUS images is proposed. Firstly In order to extract more effective features of the anatomy, using linear fitting method to select the low contrast threshold; then, a method based on the circle is proposed to describe the feature points to reduce the matching time, which greatly reduced the dimensions of the feature vector under the circumstances of guarantee the rotation invariant. Finally, using the method of module comparison to improve the matching accuracy. The experiment results show that the improved algorithm can reduce the false match rate, the total time of extraction and matching.(3) For the purpose of the effective registration of intravascular ultrasound images, an improved registration algorithm based on thin plate spline function is proposed for elastic intravascular ultrasound image registration. Firstly, the shape context algorithm is adopted to extraction and matching marked points of the outline intravascular ultrasound ROI image, then, in order to eliminate the registration distortion caused by mismatch of marked points, use the thin plate spline added smooth term for interpolating and solving the transformation parameters. Finally, the floating image is deformed by transformation parameters to finish the registration. The experimental effect showed that the proposed thin plate splines can effectively eliminate the distortion registration, and have better performance in time and accuracy compared with other registration algorithm.