Estimation of bias and variance of measurements made from tomography scans
Estimation of bias and variance of measurements made from tomography scans
复制标题
DOI：
10.1088/0957-0233/27/9/095402
复制
发表时间：
2016-09-01
影响因子：
2.4
通讯作者：
Bradley, Robert S.
中科院分区：
文献类型：
--
作者：
Bradley, Robert S.
层析成像模式被越来越多地用于量化对象的内部特征，应用范围从医学成像到材料科学研究。然而，这种测量通常在没有对其相关的方差或可信区间进行评估的情况下提出。特别是，原始扫描数据中的噪声对对重建的3D体积进行的测量的方差和偏差设置了一个基本的下限。在本文中，模拟-外推技术最初是为了统计回归而开发的，它被用来估计单次扫描测量的偏差和方差。详细讨论了该技术在X射线层析成像中的应用，并证明了该技术还可以比较自动分割策略的稳健性。
Tomographic imaging modalities are being increasingly used to quantify internal characteristics of objects for a wide range of applications, from medical imaging to materials science research. However, such measurements are typically presented without an assessment being made of their associated variance or confidence interval. In particular, noise in raw scan data places a fundamental lower limit on the variance and bias of measurements made on the reconstructed 3D volumes. In this paper, the simulation-extrapolation technique, which was originally developed for statistical regression, is adapted to estimate the bias and variance for measurements made from a single scan. The application to x-ray tomography is considered in detail and it is demonstrated that the technique can also allow the robustness of automatic segmentation strategies to be compared.