基于高频超声的人牙釉质三维重建

发布时间：2018-03-21 11:37

本文选题：高频超声　切入点：增益补偿　出处：《华南理工大学》2016年博士论文　论文类型：学位论文

【摘要】：人类牙齿健康状况越来越受到重视。很多牙齿疾病和牙齿组织中成分损失和侵蚀有关,但是在适当的治疗下可以得到控制。因此牙齿早期损伤的检测和及时治疗对防止病灶恶化是非常关键的。超声已经被广泛应用于诸如诊断、无损检测和治疗等医学领域。因为高频超声具有波长短而分辨率高的特点,更适合用于牙科医学中。虽然早在上世纪60年代超声成像技术就开始应用于牙科,但是到现在还没有发展成为有效的诊断工具。已经有很多学者和组织对三维超声成像在牙科的应用进行了尝试和探索,但是从成像效果来看并不尽如人意,有的成像系统较为复杂,大部分实验都需要对牙齿样本进行预处理,改变了样本本来面貌,降低了成像结果的实用性。针对这些缺点,本文介绍了一种新的机械扫描系统,这个系统收集人牙齿的表面结构信息和牙釉质-牙本质的边界信息并进行三维成像。中心频率为50MHz的A超换能器由一个三维控制台控制,发射并接收超声信号。对一维回波信号进行一种基于频率的增益补偿算法,然后对二维图像信号进行基于Q学习的图像去噪算法之后,可以得到较清晰的牙釉质三维图像,并不需要对牙齿做任何改变形状的处理。釉质层的测量结果和CT扫描数据的测量结果进行了比较。通过对中山大学光华口腔医学院提供的4颗成人恒磨牙分别进行的三维重建及牙釉质厚度测量的实验结果可知,本文设计的成像系统及提出的成像方法可以很好地再现牙釉质的三维轮廓,为医生提供完整的三维视图。本文的主要研究内容可分为三个方面:设计开发一个基于A超探头的三维重建系统;对一维回波信号进行基于频率的增益补偿算法;对二维超声图像进行基于Q学习的融合滤波,提高成像质量。在三维超声重建系统方面,分析总结三维超声重建系统的模块组成、各模块的技术原理及功能,对比分析研究目前存在的几种扫描方法及各自优缺点。在此基础上,设计开发了基于机械扫描的A超三维重建系统。系统主要包括:中心频率为50MHz的单晶换能器、信号发生器、三维运动台、数字信号采集卡等硬件,图像数据的采集软件和一维、二维、三维图像的处理、生成显示软件开发。在一维信号处理算法上,参考已有的增益补偿算法,将线性自回归(AR)模型应用于A超回波信号的衰减参数估计,根据估计的衰减参数对信号进行补偿,得到了较好的图像质量。在线性自回归模型基础上,将自回归参数用少量基函数的线性加权和表示,把非平稳过程转换为线性过程,提高了参数估计的精度,从而提高了衰减系数估计的准确性,进一步提高了重建图像的质量。在二维图像去噪算法上,提出了一种基于强化学习中的Q学习算法的超声二维图像融合滤波算法,利用中值滤波器、小波变换滤波器和去斑各向异性扩散滤波器共同实现二维图像融合滤波算法,对不同滤波器处理结果进行加权和,Agent在学习过程中根据一定的图像评价指标选择行动(改变各个权重的大小),直到找到最优解。主观和客观评价显示此算法具有较好效果和普遍适应性。
[Abstract]:More and more attention to the health of human teeth. Many dental disease and tooth tissue component loss and erosion, but the appropriate treatment can be obtained under control. So the detection of teeth injury in the early stage and timely treatment to prevent deterioration of the lesions is critical. Ultrasound has been widely used in applications such as diagnosis and treatment, nondestructive testing other areas of medicine. Because of high frequency ultrasound has the characteristics of short wavelength and high resolution, more suitable for dental medicine. Although in the early 60s began to use ultrasound imaging technology in dentistry, but has not yet become an effective diagnostic tool. There have been many scholars and organization of 3D ultrasound imaging in dental applications to explore, but from the imaging effect is not satisfactory, some imaging system is more complex, most experiments are needed for the teeth Sample pretreatment, changed the original appearance of the sample, reduces the usefulness of imaging results. Aiming at these shortcomings, this paper introduces a new mechanical scanning system, the boundary information of the surface structure information and the collection system of tooth enamel and dentin and three dimensional imaging. The center frequency is 50MHz A the transducer is controlled by a 3D console, transmitting and receiving ultrasonic signals. A gain compensation algorithm based on frequency of one-dimensional signal, and then the two-dimensional image signal denoising algorithm based on Q learning, you can get a clear enamel 3D image processing, does not need to do any change in the shape of teeth the comparison of measurement results. The measurement results of the enamel layer and CT scanning data. By 4 adults of Zhongshan University Guanghua School of Stomatology, the molars were the The experimental results of three-dimensional reconstruction and enamel thickness measurement shows that the 3D imaging system designed in this paper and the imaging method proposed can well reproduce the enamel, providing full 3D view for doctors. The main contents of this paper can be divided into three aspects: the design and development of a 3D reconstruction system based on A probe; gain compensation algorithm based on frequency of one-dimensional echo signal; filtering fusion Q learning based on the two-dimensional ultrasound image, improve the image quality. In 3D ultrasound reconstruction system, summarize the three-dimensional ultrasound reconstruction system module composition analysis, technical principle and function of each module, several existing scanning method and comparative analysis research the advantages and disadvantages. On this basis, the design and development of mechanical scanning system based on A 3D reconstruction system mainly includes: the center frequency is 50MHz for single crystal Can device, signal generator, three-dimensional motion, digital signal acquisition card and other hardware, image data acquisition software and the one-dimensional, two-dimensional, three-dimensional image processing, software development and display generation. In one dimensional signal processing algorithm, gain compensation algorithm with reference to the existing, the linear autoregressive (AR) model is applied to estimate the attenuation parameters A echo signal, according to the estimation of the attenuation parameters to compensate the signal, get better image quality. Based on the linear regression model, the autoregressive parameters with a small number of basis functions and linear weighted that convert non-stationary process into a linear process, improve the accuracy of parameter estimation, so as to improve the the attenuation coefficient estimation accuracy, further improve the quality of the reconstructed image. In the two-dimensional image denoising algorithm, proposed a super sound two-dimensional image learning algorithm in reinforcement learning based on Q fusion filter Using the algorithm, median filter, wavelet transform filter and speckle reducing anisotropic diffusion filter to realize two-dimensional image fusion filtering algorithm, the different results were weighted and filter processing, Agent in the process of learning to select actions according to the image evaluation index (a change of each weight size), until you find the optimal solution. The subjective and objective evaluation showed that this the algorithm has good effect and adaptability.

【学位授予单位】：华南理工大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】