【作者】 郭勇；

【导师】 邱天爽；

【作者基本信息】 大连理工大学 ， 生物医学工程， 2014， 硕士

【摘要】 手术导航系统目前已经成为计算机技术与医学领域融合的理论研究和临床应用新热点。本文研究基于超声图像引导的经皮穿刺类手术导航系统中若干关键技术。众所周知,超声成像过程通常伴随着严重的斑点噪声。其对图像细节所造成的干扰,会给进一步进行图像分割和边缘检测带来一定困难,进而影响医生进行快速准确的诊断。因此在手术导航过程中无论是进行图像配准还是图像分割,都需要先对超声图像进行去噪处理。经皮穿刺类手术的主要对象是人体软组织器官,由于软组织器官在人体呼吸运动时会产生位移,从而会影响到对术中超声图像和术前CT、MR图像的配准,同时影响穿刺的准确性。针对以上存在的问题,本文对超声图像去噪和呼吸运动的获取与建模进行了研究,主要内容如下：针对超声图像去噪问题,本文提出了基于曲率加权全变差模型的乘性噪声去噪算法。通过引入差异曲率算子,使得本文算法能够在去除乘性噪声的同时有效的保留图像边缘和细节。针对该模型使用分裂Bregman迭代方法进行求解,得到较为理想的效果。本文在Visual Studio2010和OpenCV环境下搭建了基于光学定位体外呼吸监测系统,实现了体外呼吸信号的实时检测,该系统可以检测体外三个标志点的运动,检测速率达到20帧/s。手动选取模板,利用模板匹配的方法从超声图像序列中获得体内呼吸信号,所获得的呼吸信号与体外呼吸信号有很好的相关性。利用Hessian矩阵自动提取模板,实现了从超声图像序列中自动获取体内呼吸信号。然后使用多项式模型和SVR对体内和体外呼吸运动进行建模,实验结果证明多体外单体内模型能够更好的建立体内与体外呼吸信号的相关模型。更多还原

【Abstract】 Surgical navigation system has become the current research focuses in areas of both computer technology and medical technology. In this paper we study some key technologies of ultrasound image-guided percutaneous surgical navigation system. As we all know, ultrasound imaging process is usually accompanied with severe speckle noise. The interference on the image detail caused by the noise, will bring some difficulties on further image segmentation and edge detection, and also affect doctors diagnose and efficiency. Therefore in the process of the surgery navigation, ultrasound image denoising is the first process for image registration and segmentation. The main object of percutaneous puncture operation is the tissues and organs of human, because of the displacement of the soft tissues and organs during the human respiratory movement, the registration of CT and MR before surgery and in surgery will be affected, as well the accuracy of puncture. To solve those problems, the article studies the acquirement and modeling of ultrasound denoising and respiratory movement, the main contents are:For ultrasound image denoising problem, we propose a denoising based on the curvature of the multiplicative noise weighted total variation model algorithm. Due to curvature differences is introduced, the proposed algorithm can remove multiplicative noise while preserving image edges and detail effectively. And by using Bregman iterative methods to solve the model can obtain more satisfactory results.In this paper, we build the vitro respiratory signal detection system based on optical positioning in Visual Studio2010and OpenCV environment, and achieve a real-time detection signal, this system is able to detect three landmarks, and the detection rate is up to20fps.We obtain vivo respiratory signal from ultrasound image sequences using a template matching method, it have a good correlation with in vitro respiratory signal. This paper proposes using the Hessian matrix to get in vivo respiratory signal automatically from the ultrasound image sequences. Then use the polynomial and SVR model for vitro and vivo respiratory motion modeling, experimental results show that multiple vitro signal and vivo models is better able to establish in single vitro and vivo respiratory signal correlation model.更多还原