Automated modeling, quantification, and uncertainty evaluation of branching tubular and sheet structures from 3D medical images

结合成像条件和不确定度分析，从理论上分析了由于成像过程的分辨率有限和边缘检测过程中的模糊造成的板宽测量误差。测量3D磁共振(MR)图像中的片状(或板状)解剖结构(如关节软骨和大脑皮层)的厚度通常是一项重要的诊断程序。本文的目的是研究磁共振图像中厚度测量精度的基本极限。在给定成像和后处理参数的情况下，通过理论模拟的方法推导了厚度测量精度的特征，特别是在非立方(各向异性)体素的情况下，重点讨论了板材结构取向对测量精度的影响。通过体外实验验证了理论模拟的正确性，并通过For…将点扩散效应引入到宽度测量过程中更多的是用反问题来模拟它。一种充分利用多尺度线滤波响应来估计扫描器的PSF和基于PSF的小管状结构直径的方法。该估计问题被表示为将在每个中轴点获得的多尺度响应序列与理想直线模型的预计算多尺度响应曲线进行最小二乘拟合。通过模型实验验证了该方法的有效性，并证明了该方法能够准确地测量小直径结构，而传统的基于全宽半高(FWHM)和过零边缘检测的方法高估了小直径结构。分支管状结构的提取我们开发了一种基于多方位分析的分支检测器，对于低对比度条件下的周围血管特别有效。我们将提出的分支检测器集成到一个混合血管跟踪系统中，该系统通过分析局部强度分布，根据对比评估自适应地选择两种不同的分支检测器机制。该分支检测器用于低对比度情况，而传统的基于区域生长的方法用于高对比度情况。该混合跟踪系统兼顾了高对比度情况下的稳定性和低对比度情况下的鲁棒性。通过使用腹部CT数据集进行ROC分析，表明所提出的方法是有效的。我们还使用19个腹部CT数据集证明了该方法可以从横断面图像中提取出放射科医生可观察到的90%的肝动脉。较少

Accurate quantification incorporating imaging conditions and uncertainty analysisThe measurement error of the sheet width due to limited resolution of the imaging processes and blurring involved in edge detection was theoretically analyzed. Measuring the thickness of sheet-like (or plate-like) anatomical structures, such as articular cartilages and brain cortex, in 3D magnetic resonance (MR) images is often an important diagnostic procedure. The purpose of this paper is to investigate the fundamental limits to the accuracy of thickness determination in MR images. Given imaging and postprocessing parameters, the characteristics of thickness determination accuracy are derived by means of a theoretical simulation method, focusing especially on the effect of sheet structure orientation on accuracy in the case of noncubic (anisotropic) voxels. The theoretical simulation was validated by in vitro experiments.The effects of the PSF were incorporated into the width measurement procedure by for … More mulating it by an inverse problem. A method fully utilizes multiscale line filter responses to estimate the PSF of a scanner and diameters of small tubular structures based on the PSF. The estimation problem is formulated as a least square fitting of a sequence of multiscale responses obtained at each medial axis point to the precomputed multiscale response curve for the ideal line model. The method was validated through phantom experiments and demonstrated to `accurately measure small-diameter structures which are significantly overestimated by conventional methods based on the full width half maximum (FWHM) and zero-crossing edge detection.Extraction of branching tubular structuresWe develop a branch detector using multi-orientation analysis, which is especially effective for peripheral vessels under low contrast conditions. We integrate the proposed branch detector into a hybrid vessel tracking system, which adaptively selects two different mechanisms of branch detectors based on contrast evaluations by analyzing local intensity distribution. The proposed branch detector is used in low contrast case while a conventional method based on region growing in high contrast case. The hybrid tracking system copes with both of the stability in high contrast case and the robustness in low contrast case. The proposed method was shown to be effective through ROC analysis using abdominal CT data sets. We also showed that the proposed method could extract 90% of hepatic arteries observable by a radiologist in the cross-sectional images using 19 abdominal CT data sets. Less