Jiang H, Paulsen K D, Osterberg U L, Patterson M S
Thayer School of Engineering, Dartmouth College, Hanover, New Hampshire 03755, USA.
Med Phys. 1998 Feb;25(2):183-93. doi: 10.1118/1.598179.
In this paper, an initial evaluation of our finite element based frequency-domain image reconstruction algorithm is performed for experiments where multiple millimeter-sized heterogeneities are embedded within a tissue-equivalent (optically) background medium having multicentimeter dimensions. The cases considered consist of several interesting geometry and optical property contrast combinations including (i) two different-sized targets with the same contrast at three different separation distances; (ii) two different-sized targets with different contrasts at two different separation distances; and (iii) three targets with the same and different sizes and contrasts, respectively. The reconstruction algorithm that has been used is an enhanced version of our originally developed regularized least squares approach that now includes total variation minimization, dual meshing, and spatial low-pass filtering. Quantitative measures of image quality including the size, location, and shape of the embedded heterogeneities along with errors in their recovered optical property values are presented. The results show that multiple targets can be clearly detected for all combinations of locations, sizes, and contrast levels considered, but the quantitative nature of this detection is influenced by these parameters.
在本文中，我们对基于有限元的频域图像重建算法进行了初步评估，该评估针对的是这样一些实验：在具有多厘米尺寸的组织等效（光学）背景介质中嵌入了多个毫米级的不均匀性。所考虑的情况包括几种有趣的几何形状和光学特性对比度组合，其中包括：（i）两个不同尺寸、具有相同对比度的目标，处于三种不同的分离距离；（ii）两个不同尺寸、具有不同对比度的目标，处于两种不同的分离距离；以及（iii）分别具有相同和不同尺寸及对比度的三个目标。所使用的重建算法是我们最初开发的正则化最小二乘法的增强版本，现在包括总变差最小化、对偶网格划分和空间低通滤波。给出了图像质量的定量指标，包括嵌入不均匀性的大小、位置和形状，以及它们恢复的光学特性值中的误差。结果表明，对于所考虑的位置、大小和对比度水平的所有组合，都能清晰地检测到多个目标，但这种检测的定量性质受这些参数的影响。
