Information theoretic regularization in diffuse optical tomography.
Information theoretic regularization in diffuse optical tomography.
复制标题
DOI：
10.1364/josaa.26.001277
复制
发表时间：
2009-05
期刊：
影响因子：
--
通讯作者：
C. Panagiotou;Sangeetha Somayajula;A. Gibson;M. Schweiger;R. Leahy;S. Arridge
中科院分区：
文献类型：
--
作者：
C. Panagiotou;Sangeetha Somayajula;A. Gibson;M. Schweiger;R. Leahy;S. Arridge
漫反射光学层析成像(DOT)从外部测量数据中提取介质的空间分布光学特性。恢复感兴趣的参数涉及到解决一个非线性且高度不适定的逆问题。本文利用信息论中互信息(MI)和联合熵(JE)的概念，研究了通过引入来自另一种高分辨率解剖模式的先验信息来正规化DOT的可能性。这样的泛函评估重建的光学图像和先前图像之间的相似性，同时绕过表现为两种模式中相应解剖特征的灰度值表示之间的不相称关系的多模式障碍。通过引入结构信息，我们的目的是提高解的空间分辨率和定量精度。我们从成像的角度对这一理论进行了详细的解释，并用数值模拟得到了初步的结果。此外，我们还比较了MI和JE的性能。最后，我们通过修改目标函数并将其推广到JE案例，采用了一种快速边际熵评估和优化的方法。我们演示了它在图像重建框架上的使用，并显示了显著的计算节省。
Diffuse optical tomography (DOT) retrieves the spatially distributed optical characteristics of a medium from external measurements. Recovering the parameters of interest involves solving a nonlinear and highly ill-posed inverse problem. This paper examines the possibility of regularizing DOT via the introduction of a priori information from alternative high-resolution anatomical modalities, using the information theory concepts of mutual information (MI) and joint entropy (JE). Such functionals evaluate the similarity between the reconstructed optical image and the prior image while bypassing the multimodality barrier manifested as the incommensurate relation between the gray value representations of corresponding anatomical features in the two modalities. By introducing structural information, we aim to improve the spatial resolution and quantitative accuracy of the solution. We provide a thorough explanation of the theory from an imaging perspective, accompanied by preliminary results using numerical simulations. In addition we compare the performance of MI and JE. Finally, we have adopted a method for fast marginal entropy evaluation and optimization by modifying the objective function and extending it to the JE case. We demonstrate its use on an image reconstruction framework and show significant computational savings.