本文为瑞典林雪平大学（作者：DanielForsberg）的毕业论文，共132页。

医学影像在现代医疗保健中扮演着越来越重要的角色。在医学成像中，通常需要将不同的图像相互关联起来，这是一个具有挑战性的问题，因为在不同的图像中，像素之间很少存在预定义的映射。因此，需要找到这样的映射/变换，即所谓的图像配准过程。多年来，图像配准已被证明在许多临床情况下是有用的。尽管如此，目前图像配准在临床实践中的应用是相当有限的，通常只用于图像融合。有限的应用在很大程度上是由过多的计算时间、缺乏既定的验证/度量方法以及对可变形图像配准中估计的变换可信度的普遍怀疑造成的。

本文旨在通过提出一系列的技术贡献和两个旨在提高临床效率的应用来克服限制图像配准应用的一些问题。这些贡献是在一个非参数图像配准的通用框架的背景下，使用一种称为Morphon的图像配准方法。在图像配准中，估计变换的正则化是控制配准过程的重要组成部分，本文提出了两种正则化方法，并证明了它们的适用性。尽管正则化器类似于依赖局部结构分析，但它们在实现上有所不同，其中一种实现为应用一组滤波器核，另一种实现为解决全局优化问题。此外，在估计相位差时，建议使用一组具有平行尺度的正交滤波器来驱动配准。在一组具有挑战性的图像序列上显示的一种方案，它既能提高配准过程的准确性，又能增强其鲁棒性。一般来说，计算复杂性是通过将所采用的Morphon算法移植到GPU来解决的，与单线程CPU实现相比，通过该算法可以获得38到44倍的性能改进。

建议的临床应用是基于priors的概念，它是与Morphon的初始表达形式一起制定的，表示为模型分配一组属性（局部操作符），从而指导配准过程。在这篇论文中，这是一个更进一步的步骤，在完成配准后，模型的属性被分配给病人数据。在此基础上，提出了一个应用解剖传递函数的概念，在这个应用中，不同的器官可以用不同的传递函数可视化。目前已经实现了二维切片可视化和三维体绘制。提出的第二个应用与确定描述解剖学的各种测量相关的标记被传输到患者数据中。特别地，这是适用于特发性脊柱侧凸和用于获得各种评估脊柱畸形的措施。此外，还提出了一种数据分析方案，有助于量化描述脊柱畸形的不同测量方法之间的线性相关性。

Medical imagingplays an increasingly important role in modern healthcare. In medical imaging,it is often relevant to relate difffferent images to each other, somethingwhich can prove challenging, since there rarely exists a pre-defifined mappingbetween the pixels in difffferent images. Hence, there is a need to fifind sucha mapping/transformation, a procedure known as image registration. Over theyears, image registration has been proved useful in a number of clinicalsituations. Despite this, current use of image registration in clinicalpractice is rather limited, typically only used for image fusion. The limiteduse is, to a large extent, caused by excessive computation times, lack ofestablished validation methods/metrics and a general skepticism toward thetrustworthiness of the estimated transformations in deformable imageregistration.

This thesis aimsto overcome some of the issues limiting the use of image registration, byproposing a set of technical contributions and two clinical applicationstargeted at improved clinical effiffifficiency. The contributions are made inthe context of a generic framework for non-parametric image registration andusing an image registration method known as the Morphon. In image registration,regularization of the estimated transformation forms an integral part incontrolling the registration process, and in this thesis, two regularizers areproposed and their applicability demonstrated. Although the regularizers aresimilar in that they rely on local structure analysis, they diffffer in regardto implementation, where one is implemented as applying a set of fifilterkernels, and where the other is implemented as solving a global optimizationproblem. Furthermore, it is proposed to use a set of quadrature fifilters withparallel scales when estimating the phase-difffference, driving theregistration. A proposal that brings both accuracy and robustness to theregistration process, as shown on a set of challenging image sequences.Computational complexity, in general, is addressed by porting the employedMorphon algorithm to the GPU, by which a performance improvement of 38 ∀44× is achieved, when compared to a single-threaded CPU implementation.

The suggested clinical applications are based upon the concept paint on priors, which was formulated in conjunction withthe initial presentation of the Morphon, and which denotes the notion ofassigning a model a set of properties (local operators), guiding theregistration process. In this thesis, this is taken one step further, in whichproperties of a model are assigned to the patient data after completedregistration. Based upon this, an application using the concept of anatomicaltransfer functions is presented, in which difffferent organs can be visualizedwith separate transfer functions. This has been implemented for both 2D slicevisualization and 3D volume rendering. A second application is proposed, inwhich landmarks, relevant for determining various measures describing theanatomy, are transferred to the patient data. In particular, this is applied toidiopathic scoliosis and used to obtain various measures relevant for assessingspinal deformity. In addition, a data analysis scheme is proposed, useful forquantifying the linear dependence between the difffferent measures used todescribe spinal deformities.

1. 引言

2. 局部相位和结构张量

3. 图像配准

4. 用于非参数图像配准的通用框架

5. Morphon方法

6. GPU计算

7. 基于模型的可视化

8. 特发性脊柱侧凸与脊柱畸形的评估

9. 特征脊柱：脊柱畸形的特征向量分析

10.         结论与展望

更多精彩文章请关注公众号：