基于参数化多相水平集方法的电阻抗成像算法研究

Electrical impedance tomography (EIT) is a non-invasive, radiation-free imaging technique, which comes forth in recent 30 years following morphology imaging and structure imaging. EIT is one of the important research topics in biomedical engineering. However, it is not yet fully suitable for clinical applications due to its poor spatial resolution. In this project, we intend to develop a shape-based reconstruction approach by incorporating the prior knowledge of conductivity for EIT within multiphase systems, e.g., in the application of thorax imaging, the thorax contains organs including the heart and lungs, as well as muscles, which are with different conductivity distributions. The shape to be reconstructed is implicitly represented by parametric level set (PLS) functions employing radial basis functions. The representation of the PLS function significantly reduces dimensionality of the problem and improves the condition number of inverse problem. The level set functions handle topological merging and breaking naturally during the evolution process. Further, to improve the spatial resolution of the reconstruction and achieve robust and efficient shape reconstruction, the approximation error approach will be applied into the EIT measurement model for compensating the modeling errors, e.g., errors due to discretization and inaccurate knowledge of boundary shape. On this basis, we will carry out applicability and robustness studies of the proposed approach through numerical simulations related to thorax and head imaging, and experimental studies with human thorax/head-shaped water tanks. This work allows an expansion of the parametric level set method into EIT imaging of multiphase systems and have important research value, which is expected to achieve new research findings.

电阻抗成像是当今生物医学工程学的重要研究课题之一，它作为最近三十年发展起来的新的成像技术，相比传统形态成像、结构成像具有无损、无辐射等优势。不过，目前电阻抗成像的空间分辨率相对较低，这在很大程度上限制了其临床实际应用。本项目拟针对电阻抗成像中多相体系(例如胸腔内心肺等脏器、肌肉组织等均具有不同的电导率分布)介质形状重构问题，充分利用电导率分布的先验信息，研究基于形状信息的重构算法。通过利用参数化水平集方法优越的形状拓扑结构处理能力及降维能力，结合近似误差方法优化电阻抗测量模型，以实现鲁棒高效的多相介质形状重构，从而提高电阻抗成像的空间分辨率及成像效率。在此基础上，开展胸腔、颅脑电阻抗成像仿真及含有胸腔、颅脑轮廓形状的水槽模型实验，验证算法的适用性与鲁棒性。该项目是参数化水平集方法在电阻抗成像技术中多相体系成像领域新的应用拓展，具有重要的研究价值并有望取得新的研究成果。

在本项目的资金资助和内容指导下，按期圆满完成了任务书要求内容和相应的研究工作。主要创新成果包括：(1) 创建了基于显式几何的形状重建新体系及发展了基于隐式几何的形状重建体系，攻克了隐式几何形状重建中难以直接引入几何约束的难题，揭示了显式形状先验对于改善电阻抗图像重建问题的病态特性具备优异的表现能力，显著提升了图像质量(重建图像结构相似性SSIM指标达到0.96，远优于基于平滑先验及全变分正则化方法的重建图像，其SSIM指标分别为0.76和0.80)，系列成果为开展定量化静态电阻抗成像提供了新途径；(2) 针对肺部EIT应用中动态图像低空间分辨率的技术难题，打破经典线性近似模型仅可获得定性化图像的理论局限，提出并实现了一系列应用驱动的动态图像重建算法，在国际上率先获得动物气胸、血胸的高质量三维电阻抗图像(气胸、血胸区域的体积还原率RCR指标分别为0.74和1.12，均优于传统线性差分算法RCR指标2.05和2.45)，为实现定量化高质量的动态电阻抗图像重建提供关键技术；(3)提出并成功实现了一种无需预训练的深度先验驱动的图像重建方法，通过仿真模拟、水槽模型等实验予以验证所提出方法的优越性能，该研究成果为电阻抗成像技术在病变组织特异性判断中的应用开辟了新道路。