【作者】 李伟锋；

【导师】 郁道银；

【作者基本信息】 天津大学 ， 光学工程， 2010， 博士

【摘要】 近十几年来,随着各种电子系统的应用和换能器技术的发展,超声医学诊断仪器被广泛应用于临床医学。其中多普勒血流检测超声内窥镜系统具有深度定位能力,能较好的对血流速度进行定量检测,可以更加准确地判断病变情况,得到了广泛的应用。本论文针对超声多普勒内窥成像提出全数字化的信号处理方案,实现了对内窥镜下小探头获得的多普勒微弱信号的处理,利用现场可编程门阵列(Field Programmable Gate Array, FPGA)实现信号的解调和频谱分析,并进行基于多普勒物理模型的成像实验,对超声内窥成像系统的正确性进行验证。主要内容包括:1.介绍脉冲多普勒检测技术的原理、信号模型,并针对血流自身的动力学特征及其对信号在频域上分布的影响,确定并完成超声脉冲多普勒系统的整体方案:设计脉冲多普勒系统的距离选通装置,提取目标不同深度处的血流速度信息;完成多普勒信号正交解调环节,使信号降至基频;设计FIR高通滤波器,避免血管及低速血流对信号的干扰;利用短时傅立叶变换完成频谱分析,并用动态功率谱图的形式,形象地反映血流的方向、速度等信息。2.目前有关基于内窥镜小探头的微弱超声多普勒信号检测的研究很少,本文针对超声内窥成像系统提出全数字化的多普勒回波信号处理方案。针对内窥系统超声探头体积小,回波信号微弱的特点,设计具有较高增益和较低噪声的超声信号前端接收电路。在FPGA中对微弱信号进行全数字化处理,实现正交解调与频谱分析等功能,系统具有电路匹配性好、信噪比高、处理速度快,体积小等优势。3.针对频谱多普勒系统自身的噪声和外来干扰对谱图的影响,综合小波分析在多普勒超声信号降噪和血管壁信号消除中的应用,提出一种新的超声多普勒信号分析系统设计框架。首先介绍小波变换,然后针对小波变换的缺点,给出信号的自适应描述,包括小波包和小波框架,最后介绍基于这些变换和取阈值的降噪方法及其时移不变性的获得方法,并利用基于小波框架及其时移不变特性的降噪方法对多普勒超声信号进行降噪。为了避免直接降噪复信号时,阈值非线性引起的相位扭曲,结合非固定取阈值方法,设计出一种多普勒超声信号降噪模式,并结合Matlab给出具体的实验方法和实验结果。4.搭建基于多普勒物理模型的成像实验平台,验证系统正确性。首先介绍多普勒物理模型实验平台的组成及实验方法。设计三组实验——静态血流模型、双路同向血流模型和双路异向血流模型,分析血流的速度与方向改变时,系统所得动态功率谱图的情况,证明硬件电路可以判断血流速度的快慢、辨别血流方向以及具有距离选通的功能,以此证明系统方案的可行性及正确性。本论文具有以下几个创新点:1、针对内窥镜小探头的微弱超声多普勒信号检测,设计并实现了具有较高增益和较低噪声的超声信号接收系统,在FPGA中对微弱信号进行全数字化处理,实现了正交解调与频谱分析等功能,系统具有电路匹配性好、信噪比高、处理速度快、体积小等优势。2、考虑到血流多普勒回波信号的时频特性及所携杂波的时变性、复杂性,本文应用基于小波冗余框架理论的非正交小波降噪方法,同时结合“区间能量最小”优化准则,设计了一种新型血流多普勒超声信号优化模型,通过增加适量的冗余性,来克服正交、双正交小波在信号去噪方面的不足,同时又极大地改善了小波框架的逼近空间与信号空间不匹配产生的边界畸变失真。通过数值仿真,本文设计的这种滤噪方案在信噪比改善(dB)、平均频率波形均方根误差(Hz)、谱宽波形均方根误差(Hz)等指标方面,具有比其它滤波方式更好的效果。通过实验模拟平台所得的数据处理结果验证了这一结论。3、根据血流多普勒超声信号在各尺度小波分析上的特点,针对硬阈值和软阈值算法的缺陷,采用一种新的Semisoft阈值函数,并结合二次B样条小波,通过仿真将它与软阈值、硬阈值及非负Garrote阈值进行数值对比分析,结果表明,基于Semisoft阈值法的二次B样条小波框架方法获得了最优的性能改善。更多还原

【Abstract】 During the past decades, ultrasonic medical diagnostic instruments have advanced significantly as a result of the resourceful application of electronic-system and transducer designs to the problem of noninvasive information acquisition from clinical patients. Medical Doppler ultrasonic endoscope system bases on ultrasonic endoscope system, combining the advantage of ultrasound in diagnosis with the sensitivity of Doppler to moving objects together. The Pulsed Doppler Quantitative detection can be measured by Doppler blood velocity profile. In this way, we get histology characteristic of digestive organ wall’s sections and the hemodynamics parameters on sections. The information helps us make judgments about pathological changes in upper gastrointestinal tract.This paper depicts the development of a Doppler ultrasonic endoscope imaging system based on FPGA. In order to capture the weak signal scattered from tissue, the ultrasound signal reception circuit with high gain and low noise level was designed. All-digital signal processing was utilized to process the weak signal in FPGA, and the quadrate demodulation and spectrum analysis were realized. The experiment platform based on the Doppler physical model was set up to verify the performance of the system. By comparing the sonogram under different situations, the system and signal processing method are proved reasonable and correct.This paper consists of 4 parts as follows:1. Introduce the theory and signal modal of ultrasonic pulse Doppler technique, design the integral system scheme based on hemodynamic features and its spectral influence. A distance switch was used in this system to extract information of blood flow rate. Furthermore, a quadrature demodulation module and a FIR high pass filter were adopted to depress noise from vessel and low velocity blood. Then dynamic spectral map was introduced to depict direction and velocity information of blood simultaneously.2. Currently, study of weak signal collected from small endoscope probe is still not popular. To overcome the low volume and weak amplitude problem in ultrasonic endoscope, this paper designs a totally digital process scheme for ultrasonic echo signals, and a receipt circuit with high gain and low noise. Quadrature demodulation and spectral analysis were operated in FPGA. The circuit shows advantages of well matching, high SNR, low volume and rapid process.3. A novel analysis frame on ultrasonic signal, which integrate wavelet analysis advantage on signal denoise and vessel noise removing, was developed to depress the influence of system noise and input artifacts. In this part we introduce wavelet transform firstly and then present the signal self adaptive description, which contains wavelet pack and wavelet frame. Finally we introduce the threshold denoising method based on wavelet frame transformation and time-invariant performance. To overcome the phase confusion in directly complex signal transformation, a denoise module integrated with non-fixed threshold method was designed and experiment results on Matlab was also presented.4. Construct the image experiment platform of Doppler modal and evaluate the correctness of system. In this part, constitutes and experiment method were introduced firstly. Three group experiments: static blood flow modal, dual path same direction blood flow modal, dual path different direction blood flow modal was operated in this part and the blood flow characters in these experiments were analyzed. Results of spectral map in blood velocity and direction changing shows that the circuit in this part has the ability to judge blood velocity and direction, specifically, it also has the distance switch function. Consenquently this scheme is reasonable and correct.Innovations of this paper are :1. For weak signal detection under ultrasonic endoscope micro probe, a receipt circuit with high gain and low noise was designed and realized. Quadrature demodulation and spectral analysis were operated in FPGA. The circuit of this system shows advantages of well matching, high SNR, low volume and rapid process.2. Under the consideration of the time-variance and complexity in mixed ultrasonic echo wave signals a novel signal optimization modal was designed, which combined nonorthogonal wavelet based on wavelet frame redundancy theory with interval power minimum optimization rules. Through improve the redundancy properly, this method overcome the limitation of orthogonal and biorthogonal wavelet on the denoising and reform the boundary distortion from the mismatch between wavelet frame approximate space and signal space. Results of value simulation expressed higher SNR, lower root mean square error of mean frequency waveform and spectral range waveform. Moreover, the experiment results also demonstrated this conclusion.3. Based on the multi scale wavelet character of blood ultrasonic Doppler signals, a novel semisoft threshold function was adopted in this paper to overcome the flaw of hard threshold and soft threshold algorithms. The comparison analysis with soft threshold, hard threshold and non-negetive Garrote threshold were operated combined with Quadratic B-Spline wavelet, the results showed that BWTF method based on semisoft threshold achieved the best performance.更多还原