可用于早期骨关节炎检测的三维多波长光声成像算法研究

Due to its high sensitivity to the changes in optical properties of various joint tissues and its high spatial resolution, photoacoustic tomography (PAT) is ideally suited for early detection of osteoarthritis (OA). Most of the existing reconstruction algorithms for PAT are based on the assumption of homogeneous acoustic speed, which is not suited for the highly heterogeneous joint tissues. The principal investigator of this project applied for the first time the finite element based reconstruction algorithm for detection of OA in humans, and has realized the simultaneous reconstruction of acoustic speed and other tissue functional properties. However, the reconstruction algorithm developed was based on two-dimensional (2D), and the errors caused by the 2D algorithm compared to a real three-dimensional (3D) model significantly reduced the accuracy of image reconstruction. In addition, 3D information of tissue properties was lost, resulting in accuracy for OA detection. The goal of this project is to overcome these limitations associated with the 2D algorithms by developing 3D multiscpectral quantitative PAT reconstruction algorithms coupled with a series of image enhancement schemes to minimize the impact of measurement noise, achieve improved resolution, and reduce the computational cost. We will evaluate and optimize the developed 3D reconstruction algorithms and associated enhancing schemes using tissue-mimicking phantom and clinical experiments. If successful, the proposed research will provide fast 3D PAT reconstruction algorithms for high resolution quantitative joint imaging, which will play a significant role in the clinical application of PAT for early detection of osteoarthritis.

光声成像对骨关节炎引起的光学参数的变化非常敏感，且继承了超声成像高分辨率的优点，因此很适合进行骨关节炎的早期常规诊断。但是现有的光声成像算法大多基于声速均匀的假设，不适用于声速高度非均匀的人体关节成像。申请人首次将基于有限元的多波长光声成像算法其应用在骨关节炎的临床诊断中，成功地实现了关节内声速和其它各功能参数的同时重建。然而其二维重建方式和骨关节三维模型间的误差影响了重建的精确性，并导致三维信息的缺失，因而降低了骨关节炎诊断的应用效果。本项目拟发展一套三维多波长光声算法，并采用一系列计算优化的方法，来解决三维重建中可能遇到的数据量大、噪声水平高、计算时间长等问题，然后用仿体实验和临床实验进行优化、验证和评估，从而实现一种高速、高分辨率、多功能的人体关节三维成像方式，为骨关节炎的诊断提供更准确完整的信息。本项目的研究成果将对光声成像在骨关节炎早期探测方面的应用起到明显的推动作用。

光声成像继承了光学成像方法的高组织对比度，以及超声成像的高空间分辨率和高穿透深度，相对于目前常用的X光、MRI、超声成像等常规医学成像手段，更适合作为一种早期骨关节炎的常规检测手段。其中，光声重建算法直接决定了光声成像的效果，在光声成像的推广应用中起关键作用。本项目推导并实现了一种以有限元方法为基础的三维多波长光声重建算法，来进行人手指骨关节的三维多成分重建。其中在算法中采用带权重的Tikhonov-Marquardt 正则化方法，并植入全变分最小化方法，从而提高重建的稳定性和图像质量；实现重建算法的GPU 并行化来降低计算时间代价；此外，针对之前光声成像有限元重建框架下数据前处理时不合理的直接截断的问题，提出了基于虚拟焦点的远场情况下的数据前处理方案，避免了偏离成像中心的区域的扭曲，提高了图像分辨率；本项目还提出用聚焦探测器的焦点作为虚拟点探测器来提高光声图像重建的切向分辨率，并制作了基于PVDF的点聚焦探头和基于PVDF环形阵列的聚焦探头。课题组将所发展的算法进行了仿体测试，并将其应用到人手指骨关节炎的三维光声重建中，与之前的二维光声成像算法结果进行了对比。此外，本项目还将光声成像的方法应用到了骨关节炎外的其它生物医学应用中。本项目为光声成像在早期骨关节炎诊断中的应用起到了重要的指导作用。同时，本项目所发展的一系列相关理论和算法还可以应用到光声成像的其它生物医学应用中，对光声成像的发展和应用具有重要的推动作用。