udImage registration seeks pointwise correspondences between the same or analogous objects in different images. Conventional registration methods generally impose continuity and smoothness throughout the image. However, there are cases in which the deformations may involve discontinuities. In general, the discontinuities can be of different types, depending on the physical properties of the tissue transitions involved and boundary conditions. For instance, in the respiratory motion the lungs slide along the thoracic cage following the tangential direction of their interface. In the normal direction, however, the lungs and the thoracic cage are constrained to be always in contact but they have different material properties producing different compression or expansion rates. In the literature, there is no generic method, which handles different types of discontinuities and considers their directional dependence.ududThe aim of this thesis is to develop a general registration framework that is able to correctly model different types of tissue transitions with a general formalism. This has led to the development of the eXtended Free Form Deformation (XFFD) registration method. XFFD borrows the concept of the interpolation method from the eXtended Finite Element method (XFEM) to incorporate discontinuities by enriching B-spline basis functions, coupled with extra degrees of freedom. XFFD can handle different types of discontinuities and encodes their directional-dependence without any additional constraints.ududXFFD has been evaluated on digital phantoms, publicly available 3D liver and lung CT images. The experiments show that XFFD improves on previous methods and that it is important to employ the correct model that corresponds to the discontinuity type involved at the tissue transition. The effect of using incorrect models is more evident in the strain, which measures mechanical properties of the tissues.ud
展开▼
机译：ud图像配准在不同图像中的相同或相似对象之间寻找逐点对应关系。常规配准方法通常在整个图像上施加连续性和平滑度。但是，在某些情况下，变形可能涉及不连续性。通常，取决于所涉及的组织转变的物理特性和边界条件，不连续性可以具有不同的类型。例如，在呼吸运动中，肺沿着其界面的切线方向沿着胸廓滑动。然而，在正常方向上，肺和胸腔被约束为始终接触，但是它们具有不同的材料特性，从而产生不同的压缩或扩张速率。在文献中，没有通用的方法可以处理不同类型的间断并考虑其方向依赖性。 ud ud本论文的目的是开发一种通用的配准框架，该框架能够使用A正确地模拟不同类型的组织转变。一般形式主义。这导致了扩展的自由格式变形（XFFD）注册方法的发展。 XFFD借鉴了扩展有限元方法（XFEM）的插值方法概念，通过丰富B样条基函数以及额外的自由度来合并不连续性。 XFFD可以处理不同类型的不连续点，并对其方向依赖性进行编码，而没有任何其他限制。 ud udXFFD已在数字体模，可公开获得的3D肝脏和肺部CT图像上进行了评估。实验表明，XFFD对以前的方法进行了改进，因此采用与组织过渡涉及的不连续类型相对应的正确模型非常重要。使用不正确模型的效果在应变中更为明显，该应变可测量组织的机械特性。 ud
展开▼
