Wu, Xixuan. / Thesis Ph.D. Chinese University of Hong Kong 2015. / Includes bibliographical references (leaves 115-128). / Abstracts also in Chinese. / Title from PDF title page (viewed on 26, October, 2016).

In this research three innovative registration systems were designed with the configurations of the mutual information and optimization technique: (1) mutual information combined with the downhill simplex method of optimization. (2) the derivative of mutual information combined with Quasi-Newton method. (3) mutual information combined with hybrid genetic algorithm (large-space random search) to avoid local maximum during the optimization. These automatic registration systems were evaluated with a variety of images, dimensions and voxel resolutions. Experiments demonstrate that registration system combined with mutual information and hybrid genetic algorithm can provide robust and accurate alignments to obtain a composite activation map for functional MRI analysis.

Chim, Ho Ming. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2008. / Includes bibliographical references (leaves 58-62). / Abstracts in English and Chinese. / Acknowledgements --- p.i / Abstract --- p.ii / Chapter 1 --- Introduction --- p.1 / Chapter 1.1 --- Objectives --- p.1 / Chapter 1.2 --- Motivations --- p.1 / Chapter 1.3 --- Problem Statement --- p.2 / Chapter 1.4 --- Camera Self-Localization Approaches --- p.3 / Chapter 1.4.1 --- Based on Calibration Patterns --- p.3 / Chapter 1.4.2 --- Based on Self-calibration --- p.3 / Chapter 1.4.3 --- Based on Shape and Motion --- p.4 / Chapter 1.4.4 --- The Proposed Approach - Based on Junctions --- p.5 / Chapter 1.5 --- Thesis Organization --- p.6 / Chapter 2 --- Previous Work --- p.7 / Chapter 2.1 --- Camera Self-Localization --- p.7 / Chapter 2.1.1 --- Parallel Plane Features --- p.7 / Chapter 2.1.2 --- Parallelepiped Features --- p.8 / Chapter 2.1.3 --- Single View Geometric Features --- p.8 / Chapter 2.1.4 --- Shape and Motion --- p.8 / Chapter 2.1.5 --- Other Estimation Methods --- p.9 / Chapter 2.2 --- Feature Correspondences Establishment --- p.9 / Chapter 2.2.1 --- Feature-based Object Recognition --- p.9 / Chapter 2.2.2 --- Model-based Object Recognition --- p.10 / Chapter 3 --- Preliminaries --- p.11 / Chapter 3.1 --- Perspective Camera Model --- p.11 / Chapter 3.2 --- Camera Pose from Point Correspondences --- p.15 / Chapter 3.3 --- Camera Pose from Direction Correspondences --- p.16 / Chapter 4 --- A Junction-based Approach --- p.18 / Chapter 4.1 --- Use of Junction Correspondences for Determining Camera Pose --- p.18 / Chapter 4.1.1 --- Constraints from Point Information --- p.19 / Chapter 4.1.2 --- Constraint from Direction Information --- p.21 / Chapter 4.1.3 --- Junction Triplet Correspondences --- p.22 / Chapter 4.2 --- Extraction of Junctions and Junction Triplets from Image --- p.24 / Chapter 4.2.1 --- Handling Image Data --- p.24 / Chapter 4.2.2 --- Bridging Lines --- p.25 / Chapter 4.2.3 --- """L""-junctions" --- p.26 / Chapter 4.2.4 --- """Y"" and ""Adjunctions" --- p.27 / Chapter 4.2.5 --- Junction Triplets --- p.28 / Chapter 4.3 --- Establishment of the First Junction Triplet Correspondence --- p.30 / Chapter 4.3.1 --- Ordered Junction Triplets from Model --- p.30 / Chapter 4.3.2 --- A Junction Hashing Scheme --- p.31 / Chapter 4.4 --- Establishment of Points Correspondence --- p.33 / Chapter 4.4.1 --- Viewing Sphere Tessellation --- p.33 / Chapter 4.4.2 --- Model Views Synthesizing --- p.35 / Chapter 4.4.3 --- Affine Coordinates Computation --- p.35 / Chapter 4.4.4 --- Hash Table Filling --- p.38 / Chapter 4.4.5 --- Hash Table Voting --- p.38 / Chapter 4.4.6 --- Hypothesis and Confirmation --- p.39 / Chapter 4.4.7 --- An Example of Geometric Hashing --- p.40 / Chapter 5 --- Experimental Results --- p.43 / Chapter 5.1 --- Results from Synthetic Image Data --- p.43 / Chapter 5.2 --- Results from Real Image Data --- p.45 / Chapter 5.2.1 --- Results on Laboratory Scenes --- p.46 / Chapter 5.2.2 --- Results on Outdoor Scenes --- p.48 / Chapter 6 --- Conclusion --- p.51 / Chapter 6.1 --- Contributions --- p.51 / Chapter 6.2 --- Advantages --- p.52 / Chapter 6.3 --- Summary and Future Work --- p.52 / Chapter A --- Least-Squares Method --- p.54 / Chapter B --- RQ Decomposition --- p.56 / Bibliography --- p.58

Evaluating non-rigid image registration performance is a difficult problem since there is rarely a “gold standard” (i.e., ground truth) correspondence between two images. The Non-rigid Image Registration Evaluation Project (NIREP) was started to develop a standardized set of common databases, evaluation statistics and a software tool for performance evaluation of non-rigid image registration algorithms. The goal of the work in this thesis is to build up common image databases for rigorous testing of non-rigid image registration algorithms, and compare their performance by a diverse set of evaluation statistics on our multiple well documented image databases. The well documented databases as well as new evaluation statistics have been and will be released to public research community. The performance of five non-rigid registration algorithms (Affine, AIR, Demons, SLE and SICLE) was evaluated using 22 images from two NIREP evaluation databases. Six evaluation statistics (Relative Overlap, Intensity Variance, Normalized ROI overlap, alignment of calcarine sulci, Inverse Consistency Error and Transitivity Error) were used to evaluate and compare registration performance. This thesis provides a complete and accurate reporting of evaluation tests so that others are able to get access to these results and make a comparison of registration algorithms they concerned in their specific use. Moreover, this work followed the recommendations of the Standards for Reporting of Diagnostic Accuracy (STARD) initiative to disclose all relevant information for each non-rigid registration validation test.

Lung function depends on the expansion and contraction of lung tissue during the respiratory cycle. The measurement of regional pulmonary function is of great interest and importance since many lung diseases can cause changes in biomechanical or material properties. It is also significant to study the radiation-induced changes in pulmonary function following radiation therapy. In this thesis, we propose a technique that uses four-dimensional (3D+time) CT imaging (4DCT), 3D non-rigid image registration to estimate regional lung function. Lung images reconstructed at different inflation levels are analyzed for dynamic lung function development during a breath cycle. We demonstrate local pulmonary function can be reproducibly measured using 4DCT in human subjects prior to RT. The image registration accuracy is validated using semi-automatic anatomic landmark picking system. The major contributions of this thesis include: 1) demonstrating the robustness and reproducibility of regional pulmonary function measurement using 4DCT in both sheep and human subjects, 2) developing approaches to improve the measurement reproducibility by dynamic lung volume matching and Jacobian normalization, 3) development and comparison four cubic metrics for reproducibility analysis, 4) research on time-varying lung ventilation in different breathing phases in both sheep and human subjects. Our contributions in this thesis are useful for diagnosis and assessment of lung diseases, useful for qualifying radiation induced changes in pulmonary function in irradiated and non-irradiated lung tissue.

The main function of the respiratory system is gas exchange. Since many disease or injury conditions can cause biomechanical or material property changes that can alter lung function, there is a great interest in measuring regional lung function and mechanics. In this thesis, we present a technique that uses multiple respiratory-gated CT images of the lung acquired at different levels of inflation with both breath-hold static scans and retrospectively reconstructed 4D dynamic scans, along with non-rigid 3D image registration, to make local estimates of lung tissue function and mechanics. We validate our technique using anatomical landmarks and functional Xe-CT estimated specific ventilation. The major contributions of this thesis include: 1) developing the registration derived regional expansion estimation approach in breath-hold static scans and dynamic 4DCT scans, 2) developing a method to quantify lobar sliding from image registration derived displacement field, 3) developing a method for measurement of radiation-induced pulmonary function change following a course of radiation therapy, 4) developing and validating different ventilation measures in 4DCT. The ability of our technique to estimate regional lung mechanics and function as a surrogate of the Xe-CT ventilation imaging for the entire lung from quickly and easily obtained respiratory-gated images, is a significant contribution to functional lung imaging because of the potential increase in resolution, and large reductions in imaging time, radiation, and contrast agent exposure. Our technique may be useful to detect and follow the progression of lung disease such as COPD, may be useful as a planning tool during RT planning, may be useful for tracking the progression of toxicity to nearby normal tissue during RT, and can be used to evaluate the effectiveness of a treatment post-therapy.

The main function of the respiratory system is gas exchange. Since many disease or injury conditions can cause biomechanical or material property changes that can alter lung function, there is a great interest in measuring regional lung ventilation and regional mechanical changes. We describe a technique that uses multiple respiratory-gated CT images of the lung acquired at different levels of inflation with both breath-hold static scans and retrospectively reconstructed dynamic scans, along with non-rigid 3D image registration, to make local estimates of lung tissue expansion. The degree of regional lung expansion is measured using the Jacobian (a function of local partial derivatives) of the registration displacement field. We compare the ventral-dorsal patterns of lung expansion estimated across seven phase changes and three pressure changes to a xenon CT based measure of specific ventilation in four anesthetized sheep studied in the supine orientation. Using 3D image registration to match images acquired at 50% and 75% phase points of the inspiratory portion of the respiratory cycle and 20 cm H2O and 25 cm H2O airway pressures gave the best match between the average Jacobian and the xenon CT specific ventilation respectively (linear regression, average r2=0.85 and r2=0.84). We validate the registration accuracy by 200 semi-automatically matched landmarks and both the dynamic and static scans show landmark error on the order of 2mm.

This thesis examines and identifies the problems of shape collapse in large deformation image registration. Shape collapse occurs in image registration when a region in the moving image is transformed into a set of near zero volume in the target image space. Shape collapse may occur when the moving image has a structure that is either missing or does not sufficiently overlap the corresponding structure in the target image. We state that shape collapse is a problem in image registration because it may lead to the following consequences: (1) Incorrect pointwise correspondence between different coordinate systems; (2) Incorrect automatic image segmentation; (3) Loss of functional signal. The above three disadvantages of registration with shape collapse are illustrated in detail using several examples with both real and phantom data. Shape collapse problem is common in image registration algorithms with large degrees of freedom such as many diffeomorphic image registration algorithms. This thesis proposes a shape collapse measurement algorithm to detect the regions of shape collapse after image registration in pairwise and group-wise registrations. We further compute the shape collapse for a whole population of pairwise transformations such as occurs when registering many images to a common atlas coordinate system. Experiments are presented using the SyN diffeomorphic image registration algorithm and diffeomorphic demons algorithm. We show that shape collapse exists in both of the two large deformation registration methods. We demonstrate how changing the input parameters to the SyN registration algorithm can mitigate the collapse image registration artifacts.

This thesis aims at introducing a methodology for clinical evaluation of orthodontic treatments using three-dimensional dento-maxillofacial images. Since complementary information is achieved by integrating multiple modalities, cone-beam computed tomography (CBCT) and stereophotogrammetry technologies are used to develop a methodology for tracking bone and facial skin variations over time. Our proposed methodology consists of a two-phase registration procedure. In the first phase, the multimodal images are registered using an extrinsic landmark-based registration followed by a robust Iterative Closest Points (ICP) method. In the second phase, by utilizing specific anatomical landmarks, single modal images of the skull and the mandible are registered over time using an intrinsic landmark-based registration method followed by the robust ICP algorithm. The results of registrations show that the signed error distribution of both mandible and skull registrations follow a normal distribution while all the errors fall within the CBCT precision range.

Image registration is the task of aligning two or more images into the same reference frame to compare or distinguish the images. The majority of registration methods deal with registering only two images at a time. Recently, a clustering method that concurrently registers more than two multi-sensor images was proposed, dubbed ensemble clustering. In this thesis, we apply the ensemble clustering method to deformable registration scenario for the first time. Non-rigid deformation is implemented by a FFD model based on B-splines. A regularization term is added to the cost function of the method to limit the topology and degree of the allowable deformations. However, the increased degrees of freedom in the transformations caused the Newton-type optimization process to become ill-conditioned. This made the registration process unstable. We solved this problem by using the matrix approximation afforded by the singular value decomposition (SVD). Experiments showed that the method is successfully applied to non-rigid multi-sensor ensembles and overall yields better registration results than methods that register only 2 images at a time. In addition, we parallelized the ensemble clustering method to accelerate the performance of the method. The parallelization was implemented on GPUs using CUDA (Compute Unified Device Architecture) programming model. The GPU implementation greatly reduced the running time of the method.