Thesis (MSc)--University of Stellenbosch, 2008. / Bibliography. Also available via the Internet.

Thesis (Ph. D. in Electrical Engineering)--Vanderbilt University, Dec. 2007. / Title from title screen. Includes bibliographical references.

Student Number : 0003555T - MSc (Eng) dissertation - School of Electrical and Information Engineering - Faculty of Engineering and the Built Environment / Successful radiotherapy treatments with high-energy proton beams require the accurate positioning of patients. This paper investigates computational methods for achieving accurate treatment setups in proton therapy based on the geometrical differences between a double exposed portal radiograph (PR) and a reference image obtained from the treatment planning process. The first step in these methods involves aligning the boundary of the radiation field in the PR with a reference boundary defined by the treatment plan. We propose using the generalised Hough transform (GHT), followed by an optimisation routine to align the field boundaries. It is found that this method worked successfully on ten tested examples, and aligns up to 82% of reference boundary points onto the field boundary. The next step requires quantising the patients anatomical shifts relative to the field boundary. Using simulated images, a number of intensity-based similarity measures and optimisation routines are tested on a 3D/2D registration. It is found that the simulated annealing algorithm minimising the correlation coefficient provided the most accurate solution in the least number of function evaluations.

In current practice, it is common in medical diagnosis or treatment monitoring for a patient to require multiple examinations using different imaging techniques. Magnetic resonance (MR) imaging and computed tomography (CT) are good at providing anatomical information. Three-dimensional functional information about tissues and organs is often obtained with radionuclide imaging modalities: positron emission tomography (PET) and single photon emission tomography (SPET). In nuclear medicine, such techniques must contend with poor spatial resolution, poor counting statistics of functional images and the lack of correspondence between the distribution of the radioactive tracer and anatomical boundaries. Information gained from anatomical and functional images is usually of a complementary nature. Since the patient cannot be relied on to assume exactly the same pose at different times and possibly in different scanners, spatial alignment of images is needed. In this thesis, a general framework for image registration is presented, in which the optimum alignment corresponds to a maximum of a similarity measure. Particular attention is drawn to entropy-based measures, and variance-based measures. These similarity measures include mutual information, normalized mutual information and correlation ratio which are the ones being considered in this study. In multimodality image registration between functional and anatomical images, these measures manifest superior performance compared to feature-based measures. A common characteristic of these measures is the use of the joint-intensity histogram, which is needed to estimate the joint probability and the marginal probability of the images. A novel similarity measure is proposed, the symmetric correlation ratio (SCR), which is a simple extension of the correlation ratio measure. Experiments were performed to study questions pertaining to the optimization of the registration process. For example, do these measures produce similar registration accuracy in the non-brain region as in the brain? Does the performance of SPET-CT registration depend on the choice of the reconstruction method (FBP or OSEM)? The joint-intensity based similarity measures were examined and compared using clinical data with real distortions and digital phantoms with synthetic distortions. In automatic SPET-MR rigid-body registration applied to clinical brain data, a global mean accuracy of 3.9 mm was measured using external fiducial markers. SCR performed better than mutual information when sparse sampling was used to speed up the registration process. Using the Zubal phantom of the thoracic-abdominal region, SPET projections for Methylenediphosponate (MDP) and Gallium-67 (67Ga) studies were simulated for 360 degree data, accounting for noise, attenuation and depth-dependent resolution. Projection data were reconstructed using conventional filtered back projection (FBP) and accelerated maximum likelihood reconstruction based on the use of ordered subsets (OSEM). The results of SPET-CT rigid-body registration of the thoracic-abdominal region revealed that registration accuracy was insensitive to image noise, irrespective of which reconstruction method was used. The registration accuracy, to some extent, depended on which algorithm (OSEM or FBP) was used for SPET reconstruction. It was found that, for roughly noise-equivalent images, OSEM-reconstructed SPET produced better registration than FBP-reconstructed SPET when attenuation compensation (AC) was included but this was less obvious for SPET without AC. The results suggest that OSEM is the preferable SPET reconstruction algorithm, producing more accurate rigidbody image registration when AC is used to remove artifacts due to non-uniform attenuation in the thoracic region. Registration performance deteriorated with decreasing planar projection count. The presence of the body boundary in the SPET image and matching fields of view were shown not to affect the registration performance substantially but pre-processing steps such as CT intensity windowing did improve registration accuracy. Non-rigid registration based on SCR was also investigated. The proposed algorithm for non-rigid registration is based on overlapping image blocks defined on a 3D grid pattern and a multi-level strategy. The transformation vector field, representing image deformation is found by translating each block so as to maximize the local similarity measure. The resulting sparsely sampled vector field is interpolated using a Gaussian function to ensure a locally smooth transformation. Comparisons were performed to test the effectiveness of SCR, MI and NMI in 3D intra- and inter-modality registration. The accuracy of the technique was evaluated on digital phantoms and on patient data. SCR demonstrated a better non-rigid registration than MI when sparse sampling was used for image block matching. For the high-resolution MR-MR image of brain region, the proposed algorithm was successful, placing 92% of image voxels within less than or equal to 2 voxels of the true position. Where one of the images had low resolution (e.g. in CT-SPET, MR-SPET registration), the accuracy and robustness deteriorated profoundly. In the current implementation, a 3D registration process takes about 10 minutes to complete on a stand alone Pentium IV PC with 1.7 GHz CPU and 256 Mbytes random access memory on board.

It is well known that biological structures such as human brains, although may contain the same global structures, differ in shape, orientation, and fine structures across individuals and at different times. Such variabilities during registration are usually represented by nonrigid transformations. This research seeks to address this issue by developing physically based models in which transformations are constructed to obey certain physical laws. <br /><br /> In this thesis, a novel registration technique is presented based on the physical behavior of particles. Regarding the image as a particle system without mutual interaction, we simulate the registration process by a set of free particles moving toward the target positions under applied forces. The resulting partial differential equations are a nonlinear hyperbolic system whose solution describes the spatial transformation between the images to be registered. They can be numerically solved using finite difference methods. <br /><br /> This technique extends existing physically based models by completely excluding mutual interaction and highly localizing image deformations. We demonstrate its performance on a variety of images including two-dimensional and three-dimensional, synthetic and clinical data. Deformable images are achieved with sharper edges and clearer texture at less computational cost.

It is well known that biological structures such as human brains, although may contain the same global structures, differ in shape, orientation, and fine structures across individuals and at different times. Such variabilities during registration are usually represented by nonrigid transformations. This research seeks to address this issue by developing physically based models in which transformations are constructed to obey certain physical laws. <br /><br /> In this thesis, a novel registration technique is presented based on the physical behavior of particles. Regarding the image as a particle system without mutual interaction, we simulate the registration process by a set of free particles moving toward the target positions under applied forces. The resulting partial differential equations are a nonlinear hyperbolic system whose solution describes the spatial transformation between the images to be registered. They can be numerically solved using finite difference methods. <br /><br /> This technique extends existing physically based models by completely excluding mutual interaction and highly localizing image deformations. We demonstrate its performance on a variety of images including two-dimensional and three-dimensional, synthetic and clinical data. Deformable images are achieved with sharper edges and clearer texture at less computational cost.

The main topic of this thesis is nonrigid image registration for medical applications. We start with an overview and classification of existing registration techniques. We develop a general nonrigid image registration algorithm. It uses spline functions to describe the deformation and uses multi-scale strategy to search for the optimal transformation. Then we present a new registration operator that is transitive and symmetric. We investigate the theoretical implication of these properties and apply this operator to the registration of sequences of MR cardiac images. In the second part of the thesis, two methods, one 2D and one 3D, for validation of nonrigid image registration algorithms are proposed and compared to a manual validation strategy. Both methods provide pairs of deformed images as well as corresponding true displacement fields with known accuracy. Nonrigid registration algorithms can be run on the pairs of images and their outputs can be compared to the true displacement fields that were generated manually by five observers. While these phantom validation studies do not provide physically correct deformations, they are certainly a useful way to test the algorithm's ability to recover various deformation patterns.

Includes bibliographical references (p. 51-54).

Ultrasound is a favorable tool for intra-operative surgical guidance due to its fast imaging speed and non-invasive nature. However, deformations of the anatomy caused by breathing, heartbeat, and movement of the patient make it difficult to track the location of anatomical landmarks during intra-operative ultrasound-guided interventions. While elastic registration can be used to compensate for image misalignment, its adaptation for clinical use has only been gradual due to the lack of standardized guidelines to quantify the performance of different registration techniques. Evaluation of elastic registration algorithms is a difficult task since the point to point correspondence between images is usually unknown. This poses a major challenge in the validation of non-rigid registration techniques for performance comparisons. Current validation guidelines for non-rigid registration algorithms exist for the comparison of techniques for magnetic resonance images of the brain. These frameworks provide users with standardized brain datasets and performance measures based on brain region alignment, intensity differences between images, and inverse consistency of transformations. These metrics may not all be suitable for ultrasound registration algorithms due to the different properties of the imaging modalities. Furthermore, other metrics are required for validating the registration performance on different anatomical images with large deformations such as the liver. This work presents a validation framework dedicated for ultrasound elastic registration algorithms. Quantitative validation metrics are evaluated for ultrasound images. These include a simulation technique to measure registration accuracy, a segmentation algorithm to extract anatomical landmarks to measure feature overlap, and a technique to measure the alignment of images using similarity metrics. An extensive study of an ultrasound temporal registration algorithm is conducted using the proposed validation framework. Experiments are performed on a large database of 2D and 3D US images of the carotid artery and the liver to assess the performance of this algorithm. In addition, two graphical user interfaces which integrate the image registration and segmentation techniques have been developed to visualize the performance of these algorithms on ultrasound images captured in real time. In the future, these interfaces may be used to enhance ultrasound examination. / Thesis (Master, Electrical & Computer Engineering) -- Queen's University, 2007-10-24 22:35:20.875

In current practice, it is common in medical diagnosis or treatment monitoring for a patient to require multiple examinations using different imaging techniques. Magnetic resonance (MR) imaging and computed tomography (CT) are good at providing anatomical information. Three-dimensional functional information about tissues and organs is often obtained with radionuclide imaging modalities: positron emission tomography (PET) and single photon emission tomography (SPET). In nuclear medicine, such techniques must contend with poor spatial resolution, poor counting statistics of functional images and the lack of correspondence between the distribution of the radioactive tracer and anatomical boundaries. Information gained from anatomical and functional images is usually of a complementary nature. Since the patient cannot be relied on to assume exactly the same pose at different times and possibly in different scanners, spatial alignment of images is needed. In this thesis, a general framework for image registration is presented, in which the optimum alignment corresponds to a maximum of a similarity measure. Particular attention is drawn to entropy-based measures, and variance-based measures. These similarity measures include mutual information, normalized mutual information and correlation ratio which are the ones being considered in this study. In multimodality image registration between functional and anatomical images, these measures manifest superior performance compared to feature-based measures. A common characteristic of these measures is the use of the joint-intensity histogram, which is needed to estimate the joint probability and the marginal probability of the images. A novel similarity measure is proposed, the symmetric correlation ratio (SCR), which is a simple extension of the correlation ratio measure. Experiments were performed to study questions pertaining to the optimization of the registration process. For example, do these measures produce similar registration accuracy in the non-brain region as in the brain? Does the performance of SPET-CT registration depend on the choice of the reconstruction method (FBP or OSEM)? The joint-intensity based similarity measures were examined and compared using clinical data with real distortions and digital phantoms with synthetic distortions. In automatic SPET-MR rigid-body registration applied to clinical brain data, a global mean accuracy of 3.9 mm was measured using external fiducial markers. SCR performed better than mutual information when sparse sampling was used to speed up the registration process. Using the Zubal phantom of the thoracic-abdominal region, SPET projections for Methylenediphosponate (MDP) and Gallium-67 (67Ga) studies were simulated for 360 degree data, accounting for noise, attenuation and depth-dependent resolution. Projection data were reconstructed using conventional filtered back projection (FBP) and accelerated maximum likelihood reconstruction based on the use of ordered subsets (OSEM). The results of SPET-CT rigid-body registration of the thoracic-abdominal region revealed that registration accuracy was insensitive to image noise, irrespective of which reconstruction method was used. The registration accuracy, to some extent, depended on which algorithm (OSEM or FBP) was used for SPET reconstruction. It was found that, for roughly noise-equivalent images, OSEM-reconstructed SPET produced better registration than FBP-reconstructed SPET when attenuation compensation (AC) was included but this was less obvious for SPET without AC. The results suggest that OSEM is the preferable SPET reconstruction algorithm, producing more accurate rigidbody image registration when AC is used to remove artifacts due to non-uniform attenuation in the thoracic region. Registration performance deteriorated with decreasing planar projection count. The presence of the body boundary in the SPET image and matching fields of view were shown not to affect the registration performance substantially but pre-processing steps such as CT intensity windowing did improve registration accuracy. Non-rigid registration based on SCR was also investigated. The proposed algorithm for non-rigid registration is based on overlapping image blocks defined on a 3D grid pattern and a multi-level strategy. The transformation vector field, representing image deformation is found by translating each block so as to maximize the local similarity measure. The resulting sparsely sampled vector field is interpolated using a Gaussian function to ensure a locally smooth transformation. Comparisons were performed to test the effectiveness of SCR, MI and NMI in 3D intra- and inter-modality registration. The accuracy of the technique was evaluated on digital phantoms and on patient data. SCR demonstrated a better non-rigid registration than MI when sparse sampling was used for image block matching. For the high-resolution MR-MR image of brain region, the proposed algorithm was successful, placing 92% of image voxels within less than or equal to 2 voxels of the true position. Where one of the images had low resolution (e.g. in CT-SPET, MR-SPET registration), the accuracy and robustness deteriorated profoundly. In the current implementation, a 3D registration process takes about 10 minutes to complete on a stand alone Pentium IV PC with 1.7 GHz CPU and 256 Mbytes random access memory on board.