CASIA OpenIR  > 中国科学院分子影像重点实验室
基于光源分布特性的小动物光学分子断层成像方法研究
蔡美山
Subtype博士
Thesis Advisor田捷
2021-05-17
Degree Grantor中国科学院大学
Place of Conferral中国科学院自动化研究所
Degree Discipline模式识别与智能系统
Keyword分子影像,光学分子断层成像,稀疏重建,逆向问题,凸优化
Abstract

分子影像是结合了医学、信息学、生物学、化学、材料学等多种学科的新一代特异性医学成像技术。与传统影像技术相比,其突出特征在于能够对人类和其他生命系统在分子和细胞水平上的生理变化过程进行在体、动态地可视化、表征和定量研究,目前已经广泛应用于疾病早期诊断、医学手术导航、药物疗效评估等多个领域。作为分子影像方法的重要组成部分,光学分子断层成像利用分子探针发出光学信号进行三维可视化,从而获得生物体内特定细胞分子的三维空间分布信息,为相关生理过程提供量化指标。由于灵敏度高、特异性好、成本低等优势,光学分子断层成像在过去十几年中迅速发展,在生物医学成像领域中取得了大量研究成果。然而,光子在生物组织传输过程较为复杂,特别是严重的散射效应,导致光学分子断层成像是一个高病态性问题,其成像性能面临较大挑战,表现为光源的定位精度、光源形状的恢复能力以及成像结果的鲁棒性等受到制约。

本文针对如何提高光学分子断层成像精度这一挑战性问题,研究了基于光源分布特性的光学分子断层成像方法,并在数值仿真实验和活体小动物肿瘤成像实验中进行了方法性能验证研究。首先,基于光源空间分布特性,即光源在空间中稀疏分布的先验信息,分别研究了基于贪婪策略和基于迭代优化策略的光学分子断层成像稀疏重建方法;其次,基于光谱分布特性,研究了光子在近红外二区这一新型成像谱段在不同组织中的传输特点,构建了相应的组织特异性近红外二区荧光分子断层成像前向模型,进而采用了高斯邻域融合拉索方法对其逆向求解;最后,构建了新型的近红外二区/一区荧光分子断层成像系统,为新型光学分子断层成像方法在小动物实验研究中的应用创造了条件。具体来说,本文的主要研究内容包括以下几点:

1. 基于光源空间分布特性的光学分子断层成像方法。通过将光源空间稀疏分布先验知识引入光学分子断层成像,构建了基于稀疏重建理论的光学分子断层成像方法。在稀疏重建理论中,一般包含两类重建策略:贪婪策略与迭代优化策略。本文一方面在传统贪婪策略的基础上,研究了基于前向预测正交匹配追踪的光学分子断层成像方法,通过在原子筛选中引入前向预测函数,对进入支撑集原子进行更加谨慎地评估,减小原子误选概率,从而提升重建精度;另一方面,在传统迭代优化策略基础上,研究了基于非负迭代凸优化的光学分子断层成像新方法,引入spike and slab先验模型构建非凸目标函数,将其分解为若干个凸优化子问题进行迭代求解,获得原问题的解。通过系列数值仿真实验与活体乳腺肿瘤成像实验验证了方法的有效性。

2. 基于光源光谱分布特性的光学分子断层成像方法。研究了光子在新型近红外二区谱段的传输特性。通过光子传输蒙特卡洛仿真实验和活体小鼠血管成像实验,验证了近红外二区荧光的成像优势。结合近红外二区光谱特性及扩散方程,建立了组织特异性近红外二区荧光分子断层成像前向模型;在其逆向问题求解中,提出了高斯加权邻域融合拉索方法,以提高求解精度。最后,通过单光源仿真、双光源仿真、小动物肿瘤成像、深层光源成像等实验验证了方法的有效性。

3. 新型近红外二区/一区荧光分子断层成像系统。该系统为提出的光学分子断层成像方法在小动物活体实验中的应用创造了条件。在体实验是检验光学分子断层成像方法的重要标准,也是光学分子断层成像方法研究的最终目的。本文介绍了新型近红外二区/一区荧光分子断层成像系统的组成结构及数据获取与预处理流程。基于该系统设计了多样化的小动物活体成像实验,验证了提出的光学分子断层成像新方法的活体适用性,为光学分子断层成像的转化提供新的思路。

Other Abstract

Molecular imaging is a new generation of specific medical imaging technology that combines multiple disciplines such as medicine, informatics, biology, chemistry, and materials science. Compared with traditional imaging technologies, its outstanding feature is that it can perform dynamic in vivo visualization, characterization and quantitative researches on the physiological changes of humans and other life systems at the molecular and cellular level. It has been widely used in early diagnosis of diseases, image-guided survey, evaluation of drug efficacy, etc. As an important part of the molecular imaging methods, optical molecular tomography (OMT) utilizes the molecular probes, which emit the optical signals to obtain the three-dimensional spatial distribution of in vivo specific cell molecules and to provide the quantitative indicators for the related physiological processes. With the advantages of high sensitivity, good specificity, and low cost, OMT has a rapid development in the past ten years. A large number of research results on OMT have been obtained in the field of biomedical imaging. However, the photon transmission process in biological tissues is extremely complicated, especially the severe scattering effect, which makes OMT a highly ill-conditioned problem. The imaging performance of OMT faces great challenges, which are embodied in the constraint of the location accuracy, the shape recovery capability, and the imaging robustness.  

To improve the imaging accuracy of OMT, several OMT methods are researched in this thesis based on the distribution characteristics of optical sources. The performance of these methods is validated via the numerical simulations and the in vivo tumor-bearing mice model imaging experiments. Firstly, the sparse prior of optical sources is induced in the process of OMT based on the spatial distribution characteristics. Accordingly, two sparse reconstruction methods of OMT are researched based on the greedy strategy and the iterative optimization strategy, respectively. Secondly, the photon transmission process in the novel near-infrared-II (NIR-II) window in different biological tissues is studied based on the spectral distribution characteristics. Accordingly, the tissue-specific forward model for NIR-II fluorescence molecular tomography (FMT) is established and the Gaussian weighted neighborhood fused lasso approach is developed to solve the inverse problem. Lastly, we construct a novel near-infrared-II/near-infrared-I (NIR-II/NIR-I) FMT system, creating great conditions for the applications of the researched OMT methods in the small animal studies. Specifically, the main contents of this thesis are listed as follows:

1. The OMT methods based on the spatial distribution characteristics of optical sources. Two sparse reconstruction methods for OMT are developed by inducing the sparse prior of optical sources in the process of OMT. In the theory of sparse reconstruction, there exist two types of reconstruction strategies, including the greedy strategy and the iterative reconstruction strategy. On the one hand, based on the greedy strategy, the look ahead orthogonal matching pursuit based OMT method is developed, which utilizes a look ahead function for more careful evaluation of the atoms in the support set to enhance the reconstruction accuracy. On the other hand, based on the iterative optimization strategy, the non-negative iterative convex refinement based OMT method is developed by inducing the spike and slab prior. Using this prior, a non-convex objective function is obtained, which can be decomposed into a series of convex sub-problems to approach the solution of original problem iteratively. To verify the performance of these methods, several numerical simulations and in vivo breast cancer imaging experiments are conducted.

2. The OMT method based on the spectral distribution characteristics of optical sources. Firstly, the transmission characteristics of NIR-II photons are studied with the Monte Carlo photon simulations and the mice vessel imaging experiments, which validate the advantages of NIR-II fluorescence imaging. The tissue-specific forward model of NIR-II FMT is established by combining the optical characteristics of NIR-II fluorescence and the diffusion equation. A Gaussian weighted neighborhood fused lasso approach is proposed to solve the inverse problem of NIR-II FMT. Finally, the effectiveness of this method is verified via several experiments such as single-source simulation, dual-source simulation, in vivo tumor imaging, and deep source imaging.

3. A novel NIR-II/NIR-I FMT system, which creates great conditions for the applications of the proposed OMT methods in the small-animal experiments. In vivo experiments are the important standard to evaluate the OMT methods and the ultimate goal of the OMT researches. This thesis introduces the structure and the data acquisition process of the NIR-II/NIR-II FMT system. Diversities of in vivo imaging experiments are designed to verify the in vivo applicability of the proposed OMT methods, which provides new ideas for the transformation of OMT.
 

Pages134
Language中文
Document Type学位论文
Identifierhttp://ir.ia.ac.cn/handle/173211/44742
Collection中国科学院分子影像重点实验室
Recommended Citation
GB/T 7714
蔡美山. 基于光源分布特性的小动物光学分子断层成像方法研究[D]. 中国科学院自动化研究所. 中国科学院大学,2021.
Files in This Item:
File Name/Size DocType Version Access License
毕业论文-蔡美山-提交版.pdf(10239KB)学位论文 开放获取CC BY-NC-SA
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[蔡美山]'s Articles
Baidu academic
Similar articles in Baidu academic
[蔡美山]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[蔡美山]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.