结合迭代收缩可行域的单视图多光谱生物发光断层成像  

Single-View Based Multispectral Bioluminescence Tomography with Iteratively Shrinking Permissible Region

作　　者：余景景[1] 王海玉[1] 李启越 

机构地区：[1]陕西师范大学物理学与信息技术学院,陕西西安710119

出　　处：《光学学报》2016年第12期82-89,共8页Acta Optica Sinica

基　　金：国家自然科学基金(61401264);陕西省自然科学基金(2015JM6322);中央高校基本科研业务费专项资金(GK201603025)

摘　　要：生物发光断层成像（BLT）是在已知生物组织的光学参数和生物体体表的光强分布的条件下,重建荧光光源在生物体内的分布。由于近红外光在生物组织中传输过程复杂,并且测量信息有限,BLT的光源重建问题充满挑战。为了在较少的测量下获得稳定的重建,提出了基于单视图测量的多光谱BLT重建方法,为了克服逆问题的不适定性,算法设计中结合迭代收缩可行区域策略及光源稀疏性特征。为系统评估所提出方法的光源定位能力和邻近光源的分辨能力,在匀质仿体上针对3~9mm范围内3种尺寸的光源,设计了4种深度上的单光源实验和3种间距下的双光源实验。仿真实验表明,所提出方法基于单视图测量数据,在各种光源设置下,均可以稳定准确地重建单光源目标和辨识双光源目标。Bioluminescence tomography（BLT）aims to reconstruct the internal bioluminescent source distribution with known optical properties and the measurements of light intensity on the surface of biological tissues.Due to the complexity of near-infrared photons transportation in biological tissues and the limited boundary measurements,the source reconstruction of BLT is a challenging problem.To obtain stable reconstruction with a small amount of measurements,we present a single-view based reconstruction method for multispectral BLT.In this method,we combine an iteratively shrinking permissible region strategy with sparsity-inducing regularization technique to deal with the high ill-posedness of BLT inverse problem.Simulations based on homogeneous numerical phantom are designed to systematically assess the performance of the proposed method in terms of localization accuracy and the ability to resolve two neighboring targets.We investigate single spherical sources with 3sizes at 4different depths,and double sources with 3source separation distances varying from 3to 9mm.Simulation results show that based on single view measurement the proposed reconstruction method can yield stable reconstruction with an average0.5mm accuracy in single source cases and resolve two neighboring sources with 3mm separation distance.

关 键 词：成像系统 多光谱生物发光断层成像 稀疏重建 迭代算法 逆问题 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论] Q632[自动化与计算机技术—计算机科学与技术]