【作者】 黄凯；

【导师】 王威琪； 他得安；

【作者基本信息】 复旦大学 ， 生物医学工程， 2009， 博士

【摘要】 骨质疏松症是影响公众健康的重要问题之一。患有骨质疏松时，松质骨的骨矿密度(BMD)下降、骨微结构退化。相比于双能X线吸收法(DXA)、定量CT(OCT)等诊断方法，超声诊断具有无电离辐射、便捷、速度快、价格低廉等优势。传统的超声透射法使用双探头获得人体跟骨处的超声透射信号，以参数宽带超声衰减(BUA)和超声传导速度(SOS)反映BMD信息。而超声背散射方法使用单探头，易于探测跟骨以外的骨组织(例如脊骨、腕骨等)，且超声背散射信号中含有全部的骨微结构信息，因而超声背散射法诊断骨质疏松的技术越来越受到研究者的重视。本文以松质骨中的超声背散射信号分析、松质骨微结构参数估计、超声背散射参数成像为主线，主要研究了以下内容：1．在介绍松质骨超声背散射模型的基础上，实验分析了单圆柱仿体的超声背散射频率特性，从而验证单圆柱模型解释松质骨中超声背散射的可行性，为超声背散射信号中含有松质骨的微结构信息提供理论依据。2．在松质骨微结构中，主要的散射元为骨小梁，平均骨小梁间距(MTBS)是松质骨微结构的重要参数之一，患有骨质疏松时MTBS会增大。因此，在以上研究的基础上，本文对直接从超声背散射信号中估计MTBS的算法进行研究。(1)提出适合松质骨超声背散射的仿真系统，在系统中将影响MTBS估计的组织特性具体为四个主要参数：MTBS、间距的标准差、弥散散射与规则散射的回波能量比、以及所有散射回波与噪声的能量比。(2)提出两种从超声背散射信号中估计MTBS的算法：频率域的改进的倒谱算法和时间域的简易反向滤波跟踪(SIFT)算法。通过仿真、仿体、离体牛胫骨、在体人体跟骨实验将提出的算法与已有的常用算法进行比较。其中，仿真信号由本文提出的适合松质骨超声背散射的仿真系统获得；仿体、离体牛胫骨、在体人体跟骨信号由本文搭建的背散射系统获得。实验结果表明，改进的倒谱方法比传统倒谱方法有更大的有效MTBS估计范围，并对骨小梁间距的变化、弥散散射和噪声有更强的鲁棒性；而提出的时间域的SIFT算法不仅比基于频率域的方法(倒谱法和二次变换法)鲁棒性更强，而且稳定性更好。3．由于松质骨的各向异性，松质骨微结构中各点的声特性并不一致，而超声背散射参数成像可直观给出一个区域的松质骨微结构信息。因此，本文对松质骨超声背散射参数成像进行了研究。(1)提出基于超声背散射法的超声参数成像方法，对四个超声参数(声特性阻抗(Z_b)、表观背散射系数(BC)、表观积分背散射系数(AIB)、和频谱偏移量(SMS))进行估计，以离体牛胫骨为样本实现了这四个参数成像；(2)对四个超声参数与微CT获得的松质骨微结构参数(平均骨小梁厚度(Tb．Th)、平均骨小梁间距(Tb．Sp，微CT使用的MTBS术语)、骨体积分数(BV／TV)、骨表面积体积比(BS／BV)和骨材料密度(BD))作了相关性分析。结果表明，Zb、BC、AIB、SMS参数图像从不同角度给出了松质骨物理特性的分布结果：①Z_b参数与骨微结构无较强相关性(Tb．Th，r=-0.233；Tb．Sp，r=0.257)，但与骨材料密度(BD)的相关性较强(r=0.448，p＜0.05)。Z_b参数图像表征的是不同位置处松质骨表面的物理特性。②BC和AIB参数与Tb．Sp有较强的负相关关系r=-0.513～0.596，p＜0.01)，而与Tb．Th呈现正相关(r=0.265～0.339)，但相关性并不显著，与BV／TV和BS／BV的相关性也并不明显。BC和AIB参数图像表征的是不同位置处松质骨内部骨小梁网状结构的散射强度。③SMS参数与Tb．Th、Tb．Sp、BV／TV、BS／BV和BD均有较强相关性(Tb．Th，r=-0.699，p＜0.01；Tb．Sp，r=0.477，p＜0.05；BV／TV，r=-0.675，p＜0.01；BS／BV，r=0.663，p＜0.01；BD，r=0.663，p＜0.01)。SMS参数图像表征的是不同位置处松质骨结构引起超声能量损失的程度。④BC、AIB参数值的标准差与骨结构信息有较为显著(p＜0.01)的相关性(Tb．Th，r=-0.550～-0.645；Tb．Sp，r=0.405～0.627；BV／TV，r=-0.572～-0.668；BS／BV，r=0.513～0.640)；SMS参数值的标准差也有类似的结果(Tb．Th，r=-0.720；Tb．Sp，r=0.771；BV／TV，r=-0.754；BS／BV，r=0.802)，这说明BC、AIB和SMS参数的空间变化量可能有助于骨质疏松的诊断。以上研究结果为超声背散射法评价松质骨状况和诊断骨质疏松提供了一定的理论依据。更多还原

【Abstract】 Osteoporosis （OP） has been becoming a serious public issue. Cancellous bonewith this chronic bone disease has low bone mineral density （BMD） and deterioratedbone microstructure. Compared with DXA, QCT and other diagnostic methods,ultrasound diagnostic technique offers advantages including lack of ionizing radiation,portability, speediness and low cost. Conventional ultrasound diagnostic technique isultrasonic transmission method, in which two transducers are used to acquireultrasonic transmission signals at human calcaneous bone and then ultrasonicparameters （broadband ultrasound attenuation （BUA） and speed of sound （SOS）） arecalculated to represent BMD. The novel ultrasound technique is ultrasonic backscattermethod, which uses only one transducer, thus enabling direct analysis of somecommon fracture sites such as spine and wrist other than the cancaneus. In addition,ultrasonic backscatter can provide additional microstructure information. Thus,currently, researchers have been becoming interested in investigating the potential ofultrasonic backscatter for osteoporosis diagnosis.This thesis was organized with the thought of starting from ultrasonic backscattersignal analysis to bone’s micro-structural parameter estimation, and ending withultrasonic backscatter parametric imaging of cancellous bone. The contents aresummarized as follows:1. Based on the introduction of ultrasonic backscatter theoretical model ofcancellous bone, single-cylindrical phantom experiments were conducted to analyzethe frequency-dependence of ultrasonic backscatter, thus verifying the ability ofsingle-cylindrical model to explain ultrasonic backscatter. This can help providetheoretical foundation for the concept that ultrasonic backscatter signals can offeradditional microstructure information of cancellous bone.2. In cancellous bone, trabeculae are considered as major scatterers, and the meanspace of these trabeculae （Mean trabecular Bone Spacing, MTBS） is an importantparameter to characterize cancellous bone. This thesis conducted some researches ondirectly extracting MTBS information from ultrasonic backscatter signals usingseveral signal processing algorithms.（1） A simulation system is proposed to generate ultrasonic backscatter simulatedsignals for cancellous bone. In the system, the characterization of quasi-periodicmicrostructures of cancellous bone can almost be represented by four parameters: mean trabecular bone spacing （MTBS）, regular+diffuse scattering to white noise SNR(SNR_wn), standard deviation of regular spacing （JITTER）, and amplitude ratio ofdiffuse scatterers to regular ones （A_d）.（2） Two algorithms are presented to estimate MTBS from ultrasonic backscattersignals: a frequency-domain based technique named improved AR cepstrum and atime-domain based technique named Simplified Inverse Filter Tracking （SIFT）algorithm. Comparisons were conducted to test the performance of the two proposedmethods using the backscatter data from simulation system, phantom, bovinetrabeculae in vitro, and human calcaneous bone in vivo. The simulated data weregenerated from the proposed simulation system, and the real backscatter data fromphantom and bone samples were acquired by a backscatter measurement system.Experimental results demonstrated that the improved AR cepstrum has a larger MTBSestimation range, and more robustness to the randomness in trabecular spacing,diffuse scattering and noise than the conventional AR cepstrum. On the other hand,the proposed SIFT algorithm outperformed the AR cepstrum technique in all casesand compared well with the quadratic transformation （QT） technique. In addition tostronger robustness, SIFT algorithm possesses better stability than thefrequency-based methods such as AR cepstrum and QT technique.3. Acoustic properties are not consistent at different positions of cancellous bonedue to its high anisotropy, while ultrasonic parametric imaging can provide adirect-view of the distribution of bone microstructure information. Thus, researcheswere conducted on ultrasonic parametric imaging of cancellous bone in this thesis.（1） Ultrasonic parametric imaging techniques were proposed based on ultrasonicbackscatter method. Four parameters were estimated from backscattered signals,including acoustic characteristic impedance （Z_b）, apparent backscatter coefficient（BC）, apparent integrated backscatter coefficient （AIB） and spectrum maximum shift（SMS）, and parametric images with these four parameters were constructed.（2） Correlations were studied between the four parameters and bonemicro-structural parameters obtained from aμ-CT, which are mean trabecularthickness （Tb.Th）, mean trabecular spacing （Tb.Sp, the terminology of MTBS used asμ-CT result）, bone volume/total volume ratio （BV/TV）, bone surface/volume ratio（BS/BV）, and bone material density （BD）. Results demonstrated that Z_b, BC, AIB,SMS parametric images provide the physical property distributions of cancellous bonein different angles: ①Parameter Z_b has no strong correlation with bone microstructure （Tb.Th,r=-0.233; Tb.Sp, r=0.257）, but a mediate correlation with bone material density （BD）（r=0.448, p＜0.05）. The Z_b parametric image can show the physical properties relatedto the upper surface of cancellous bone at different locations.②Parameter BC and AIB have negative correlations with Tb.Sp （r=-0.513～-0.596, p＜0.01） and increase with Tb.Th （r=0.265～0.339）, but the correlation withTb.Th is not significant as well as the correlations with BV/TV and BS/BV. The BCand AIB parametric images can offer the distribution of backscatter strengths of innerbone microstructure.③Parameter SMS has mediate and significant correlations with all theparameters obtained fromμ-CT （Tb.Th, r=-0.699, p＜0.01; Tb.Sp, r=0.477, p＜0.05;BV/TV, r=-0.675, p＜0.01; BS/BV, r=0.663, p＜0.01; BD, r=0.663, p＜0.01）. The SMSparametric image displays the distribution of ultrasonic energy lost when ultrasoundpenetrates into and propagates inside bone microstructure.④The spatial variation of BC and AIB values was significantly （p＜0.01） relatedwith bone micro-structural parameters （Tb.Th, r=-0.550～-0.645; Tb.Sp, r=0.405～0.627; BV/TV, r=-0.572～-0.668; BS/BV, r=0.513～0.640）, and the spatial variation ofSMS has the similar result （Tb.Th, r=-0.720; Tb.Sp, r=0.771; BV/TV, r=-0.754;BS/BV, r=0.802）, which indicate that the spatial variation of parameters BC, AIB andSMS may help characterize bone quality and osteoporosis diagnosis.Research results in this thesis provide theoretical support for ultrasonicbackscatter in evaluating the status of cancellous bone and diagnosing osteoporosis.更多还原