针灸疗效评估中激光散斑衬比图像稳定性研究

激光散斑衬比成像技术可以为中医理疗提供无创、高效的疗效评估手段,但目前该评估方法的稳定性有待提高。基于能产生稳定散斑衬比的近红外激光光源,对针灸疗效进行散斑成像实验并获得原始实验数据,分析对比Roll算法与基于此算法加入时间抽样形成新的衬比处理RT(Roll with time-sampling)算法的数据处理结果。从衬比随时间分布的规律可以看出,RT算法处理所得的衬比分布更加均匀,消除了部分噪声干扰,因此RT算法能够提升激光散斑衬比分布的稳定性。     

一种联合光声与激光散斑的多模态成像系统研究

多模态成像技术是近年生物医学成像领域的研究方向之一,它能够结合不同模式的成像,解决单一成像模式的不足。论文设计并搭建了一套基于光声成像与激光散斑成像的多模态成像系统,通过对头发光声成像和对血流样品激光散斑成像的模拟实验,验证了系统的可行性,测得光声成像分辨率为40 μm,激光散斑成像血流速度分辨率为0.2 mm/s。通过在体实验实现了对小鼠耳朵血管损伤的连续7天检测,分析血管损伤和修复情况。该系统有效的结合了高分辨率光声成像技术与激光散斑血流成像技术,对新型的多模态光声成像系统研究具有重要意义。     

激光散斑血流成像系统中的光源相干性

针对激光散斑血流成像系统中光源相干性对散斑衬比的影响进行了研究,提出了一种基于混合遗传算法的散斑图像恢复方案。首先,分析非完全相干光用于激光散斑血流成像系统时的劣势,说明混合遗传算法恢复被低相干性光源破坏的散斑图像的原理;使用对比度与噪声比来衡量由两种光源成像得到的静态白盘散斑图像质量;最后,将算法分别用于静态白盘散斑图像和真实的手背血流检测场景。实验结果表明,相干性较好的单纵模激光器成像得到的散斑图像,较之于相干性较差的半导体激光器成像得到的散斑图像,对比度与噪声比高出67.2%;将算法用于质量较差的散斑图像后,其对比度与噪声比提高了57.7%;将算法用于手背血流检测场景,进行恢复运算后,验证了手指指尖的血流灌注量平均高于手指关节30%的临床事实。将本算法用于激光散斑血流成像系统,可以有效消除非完全相干光给散斑衬比分析造成的影响,获得质量更好的血流图像。本算法的推广使用可以提高激光散斑血流仪器在光源选择方面的灵活性。     

激光散斑血流成像系统及其在生物医学工程专业实验教学中的应用

实验教学有利于加深生物医学工程专业学生对交叉学科知识的理解,提高学生的综合素质。论文探讨了激光散斑血流成像系统在实验教学中的应用,详细介绍了激光散斑血流成像系统的原理和组成。论述了使用该系统开展生物光学光路设计、数字图像采集、图像重建算法、虚拟仪器软件设计等实验内容和针对不同类型实验的教学组织形式。最后给出演示实验和动物血流生理实验两个具体实施案例。研究表明,激光散斑血流成像系统是一个高效的综合实验平台,在生物医学工程专业实验教学中具有重要的作用。     

监测区域性流速分布的激光散斑成像系统

建立了一套激光散斑成像系统 ,对该成像系统的性能参数进行了分析 ,并利用该系统对区域性的流速分布进行监测 ,结果表明 ,该系统能在不需扫描的情况下对区域性的流速分布进行有效监测 ,具有较高的时间分辨率 (2 5 m s)和空间分辨率(取决于系统的放大倍数 ,可达几个 μm) ,对较慢的速度变化敏感。该系统为监测区域性流速分布提供了一种高时—空分辨率的实时动态手段     

激光散斑血流成像对中医理疗功效的检测

为了获得更加准确的中医理疗功效检测结果,根据中医经络的气血和人体皮肤微循环密切相关的理论,提出了一种基于散斑血流成像的检测方法。对检测过程中由于被测对象身体抖动等因素产生的图像信噪比下降问题进行研究和改进,并通过软管模拟实验对成像系统的稳定性进行验证。然后,对10名健康志愿者的内关穴或外关穴进行艾灸实验,以血流灌注指数为检测指标,采用散斑血流成像系统对手部不同经络的穴位及非穴位检测点进行检测。实验结果表明:艾灸后,在艾灸点所在经络上的穴位检测到的血流灌注指数增加比其它穴位点高20%。激光散斑血流成像系统能够对理疗过程中皮肤血流变化予以大视角的成像,此方法在中医理疗功效检测中具有有效性。     

基于滤光轮双相机系统的高光谱分辨率成像

滤光轮光谱成像系统在光谱成像领域应用广泛,空间分辨率高,但是光谱分辨率较低。针对这一问题,提出了基于滤光轮双相机系统的高光谱分辨率成像,设计了一种基于插分补偿的多光谱计算重构方法,实现系统的高光谱分辨率、高空间分辨率成像。首先利用滤光轮双相机成像系统采集多光谱图像以及RGB图像,然后从多光谱图像获取离散的光谱响应曲线,最后根据RGB三通道数据与光谱高维数据之间的映射关系以及能量守恒定理,进行光谱响应曲线的插分补偿并实现高光谱分辨率成像。实验结果表明,本文方法能够在保持空间分辨率的情况下,高效地实现光谱分辨率为5 nm甚至更高光谱分辨率的成像,重建结果与真实值的均方根误差为0.017 1,具有较好的准确性和鲁棒性。     

临床X射线相衬成像研究进展

X射线相衬成像在提高人体软组织成像的衬度分辨率及空间分辨率研究中具有潜在优势,是一项有着广阔发展前景的技术。它利用空间相干X射线投过物体后携带的相位信息进行成像,在临床上适用于软组织物体,可以减少吸收剂量,进而减轻病人的辐射损伤。本文详述了当今几种典型的X射线相衬成像方法,包括X射线干涉相衬,衍射增强相衬以及类同轴相衬的成像原理以及发展历程。通过对比表明了几种成像方法的优缺点。最后对临床X射线相衬成像有待解决的问题作了分析。     

基于反投影的MIT动态图像重建方法研究

磁感应断层成像(magnetic induction tomography,MIT)是一种无创、非接触的新型医学成像技术,图像重建算法是实现MIT快速、精确成像的关键.提出一种改进的反投影图像重建算法,首先根据成像区域的磁场分布,由磁力线确定反投影路径,降低了直线反投影用于磁场成像的定位误差;其次根据MIT电磁关系推导,构建了边界检测数据的修正模型,据此对边界相位差数据进行修正处理,进一步提高了重建图像定位准确性;最后分别对成像区域内扰动目标电导率大小变化及位置变化2种情况,构建了序列重建图像,对该图像序列联合分析获取纵向阻抗变化信息,反映了成像体随时间变化的动态信息.实验结果表明该算法具有重建速度快、定位准确的特性,能够准确反映成像区域内部电导率变化,结合序列图像联合分析实现MIT动态成像.     

球面弯晶成像的优化设计及实验测试

为满足"神光Ⅱ"升级激光装置实验的激光等离子体高空间分辨和窄带宽的精密成像诊断需求,本文研究球面弯晶单色成像技术,并对其进行优化设计及实验验证。首先,使用光线追迹方法分析了成像像距、布拉格角、背光源尺寸等成像系统参数对球面弯晶成像性能的影响,分析了设计参数的变化对成像空间分辨能力的影响,并在X射线管上开展了测试验证实验。研究结果表明:二维成像时,在等分辨成像像距处可以获得较好的成像结果,增大布拉格角可以抑制像差进而提高成像质量,更小的背光源尺寸也有助于提高成像的空间分辨率。测试实验获得了清晰的网格背光成像,成像结果与预期相符,空间分辨率优于14μm。这一工作有助于球面弯晶成像系统的优化设计及其在激光等离子体精密诊断中的应用。     

Portable laser speckle perfusion imaging system based on digital signal processor

The ability to monitor blood flow in vivo is of major importance in clinical diagnosis and in basic researches of life science. As a noninvasive full-field technique without the need of scanning, laser speckle contrast imaging (LSCI) is widely used to study blood flow with high spatial and temporal resolution. Current LSCI systems are based on personal computers for image processing with large size, which potentially limit the widespread clinical utility. The need for portable laser speckle contrast imaging system that does not compromise processing efficiency is crucial in clinical diagnosis. However, the processing of laser speckle contrast images is time-consuming due to the heavy calculation for enormous high-resolution image data. To address this problem, a portable laser speckle perfusion imaging system based on digital signal processor (DSP) and the algorithm which is suitable for DSP is described. With highly integrated DSP and the algorithm, we have markedly reduced the size and weight of the system as well as its energy consumption while preserving the high processing speed. In vivo experiments demonstrate that our portable laser speckle perfusion imaging system can obtain blood flow images at 25 frames per second with the resolution of 640 × 480 pixels. The portable and lightweight features make it capable of being adapted to a wide variety of application areas such as research laboratory, operating room, ambulance, and even disaster site.     

投影散斑相关方法测量芯片翘曲度

本文采用投影散斑相关法对芯片翘曲度进行了测量。利用单个CCD和单个投影装置多次扫描的方法，得到若干幅具有部分重叠区域的小图像，然后通过配准和拼接获得全景图像，实现整体测量。在拼接区标定若干控制点，然后以这些控制点为基准进行坐标变换实现重叠区域的配准和拼接，以提高分辨率、实现全场测量，得到芯片相对于标准平面的面内位移和形变。通过三角法标定，获得了离面位移和形变，从而完成了对手机屏幕芯片翘曲度的测量。与牛顿环测量法相比较，结果证明了投影散班方法测量芯片翘曲度的精度和可行性。     

Tracking quasi-stationary flow of weak fluorescent signals by adaptive multi-frame correlation

We have developed a novel cross-correlation technique to probe quasi-stationary flow of fluorescent signals in live cells at a spatial resolution that is close to single particle tracking. By correlating image blocks between pairs of consecutive frames and integrating their correlation scores over multiple frame pairs, uncertainty in identifying a globally significant maximum in the correlation score function has been greatly reduced as compared with conventional correlation-based tracking using the signal of only two consecutive frames. This approach proves robust and very effective in analysing images with a weak, noise-perturbed signal contrast where texture characteristics cannot be matched between only a pair of frames. It can also be applied to images that lack prominent features that could be utilized for particle tracking or feature-based template matching. Furthermore, owing to the integration of correlation scores over multiple frames, the method can handle signals with substantial frame-to-frame intensity variation where conventional correlation-based tracking fails. We tested the performance of the method by tracking polymer flow in actin and microtubule cytoskeleton structures labelled at various fluorophore densities providing imagery with a broad range of signal modulation and noise. In applications to fluorescent speckle microscopy (FSM), where the fluorophore density is sufficiently low to reveal patterns of discrete fluorescent marks referred to as speckles, we combined the multi-frame correlation approach proposed above with particle tracking. This hybrid approach allowed us to follow single speckles robustly in areas of high speckle density and fast flow, where previously published FSM analysis methods were unsuccessful. Thus, we can now probe cytoskeleton polymer dynamics in living cells at an entirely new level of complexity and with unprecedented detail.     

基于时空张量融合的人体骨架行为自适应识别方法

针对人体行为的空间复杂性和时间差异性问题,提出了一种基于时空张量融合的人体骨架行为自适应识别方法。 首先 充分利用人体行为骨架序列的帧内空间关系和帧间时间关系,构建相邻帧时空特征张量;其次通过计算相邻帧时空特征张量的 差异性获取关键相邻帧时空特征张量并组成行为时空特征张量;之后利用行为时空特征张量的空间特征差异和多尺度时间卷 积构建行为时空特征张量自适应注意力机制,完成行为时空特征融合;最后,使用深度随机配置网络根据行为时空特征融合张 量识别人体行为。 使用 NTU RGB-D 数据集进行实验仿真,识别准确率达到 84. 57% ,并且设计相应的系统进行实际应用验证, 结果表明本文所提方法是一种适合应对人体行为空间复杂性和时间差异性问题的人体行为识别方法。     

Analysis of blood flow covering a wide region of velocity in laser speckle image sensing

Imaging technique based on bio-speckles is useful for visualizing blood flow distribution. We have proposed so far temporal contrast factor (TCF) and inversed temporal contrast factor (ITCF) for imaging of slow and fast blood flows, respectively. However, selection of TCF and ITCF depends on user’s experience. Additionally, these parameters are significantly influenced by the speckle contrast and random noises. In the present study, we propose a method for automatic selection of a proper parameter for imaging blood flow in an unknown velocity region with reduced effects of speckle contrast and random noises. Experiments are conducted for a sample model using ground-glass plates, human wrist, and human finger to confirm the feasibility of the proposed method.     

基于LSTM-RF模型的城市道路交通流预测

交通流预测作为信号协调和出行时间预测等任务的基础,成为了交通领域的研究点。对于交通流预测问题,研究人员提出了多种方法,但这些方法大多只使用交通流数据的时域信息进行交通流预测,忽略了空间相关性对于预测目标路段流的影响,导致预测精度不理想。基于组合模型的思想提出了一种称为LSTM-RF的交通流预测模型。在交通流预测过程中,首先使用LSTM模型提取预测目标路段的时序特征,再将其预测值与采集的相邻上下游路段信息同时作为随机森林模型的输入特征,进行交通流时空相关性分析,获得最终的预测结果。并通过贵阳市车牌识别系统采集的城区132条路段的交通流数据进行实验验证。结果表明:该方法在预测精度上优于单一模型,并且预测误差相比单一模型有明显减少。     

基于超声造影剂谐波图像的流场测量技术

本文研究了基于超声造影剂的血流运动场估计与显示技术,通过在血液中注入超声造影剂作为示踪粒子,对造影谐波图像进行时域相关处理,可得到成像部位的二维流场分布图。相对于常规Ｄｏｐｐｌｅｒ方法中用Ｄｏｐｐｌｅｒ回波信号的频偏计算流速值,该技术可直接从超声图像提取与夹角无关的流速矢量信息。本文通过流动模型验证该方法,浸入超声水槽中的乳胶管中流动着血液替代品,沿水流方向进行超声成像,对实验所得的造影后的Ｂ超图像以及谐波图像,用一种节省计算量的多尺度相关算法进行处理,并相互比较。结果显示,谐波图像相对Ｂ超基波图像具有更高的信杂比,从根本上解决了基波图像低信噪比对时域相关测量精度的限制,可以得到与夹角无关的二维血流场分布图,该方法是医学和有关工业领域中超声流场测量的一种有效的方法。     

基于散斑干涉的相位测量技术研究

本文对几种基于散斑干涉方法的相位测量技术进行了研究与比较,包括时间相移法、空间相移法、空间载波相移法以及空间载波傅立叶变换法,分析和比较了这些方法的优缺点,并对时间相移法和空间载波傅立叶变换法进行了试验研究。     

Differences between time domain and Fourier domain optical coherence tomography in imaging tissues

It has been numerously demonstrated that both time domain and Fourier domain optical coherence tomography (OCT) can generate high‐resolution depth‐resolved images of living tissues and cells. In this work, we compare the common points and differences between two methods when the continuous and random properties of live tissue are taken into account. It is found that when relationships that exist between the scattered light and tissue structures are taken into account, spectral interference measurements in Fourier domain OCT (FDOCT) is more advantageous than interference fringe envelope measurements in time domain OCT (TDOCT) in the cases where continuous property of tissue is taken into account. It is also demonstrated that when random property of tissue is taken into account FDOCT measures the Fourier transform of the spatial correlation function of the refractive index and speckle phenomena will limit the effective limiting imaging resolution in both TDOCT and FDOCT. Finally, the effective limiting resolution of both TDOCT and FDOCT are given which can be used to estimate the effective limiting resolution in various practical applications.     

Technique for the measurement of spatial vorticity distributions

A method for measuring mean spatial vorticity content has been described for flows where the primary convective velocity (free stream) is larger than the mean crossflow components. Employing an 'X'-geometry hot-wire probe and linearized anemometer system in conjunction with a precision probe traversing mechanism, this relatively simple technique can be used to evaluate the circulation associated with various spatial contours. Errors associated with the technique are computed exactly and in simple linearized forms to illustrate its validity in a wide range of flows. A priori knowledge of the flow character and resultant careful selection of the integration paths for a given flow situation is observed to result in minimal measurement errors. Application to the problem of an isolated trailing vortex is presented as an example. Experiments at the trailing edge of a lifting wing confirm the accuracy of the method when compared with existing theories and other experimental data. The method can be extended to periodically driven unsteady flows using phase-lock techniques.