Our motivation is to develop an automated technique for the segmentation of sub-cortical human brain structures from MR images. To this purpose, models of shape-and-appearance are constructed and fit to new image data. The statistical models are trained from 317 manually labelled T1-weighted MR images. Shape is modelled using a surface-based point distribution model (PDM) such that the shape space is constrained to the linear combination of the mean shape and eigenvectors of the vertex coordinates. In addition, to model intensity at the structural boundary, intensities are sampled along the surface normal from the underlying image. We propose a novel Bayesian appearance model whereby the relationship between shape and intensity are modelled via the conditional distribution of intensity given shape. Our fully probabilistic approach eliminates the need for arbitrary weightings between shape and intensity as well as for tuning parameters that specify the relative contribution between the use of shape constraints and intensity information. Leave-one-out cross-validation is used to validate the model and fitting for 17 structures. The PDM for shape requires surface parameterizations of the volumetric, manual labels such that vertices retain a one-to-one correspondence across the training subjects. Surface parameterizations with correspondence are generated through the use of deformable models under constraints that embed the correspondence criterion within the deformation process. A novel force that favours equal-area triangles throughout the mesh is introduced. The force adds stability to the mesh such that minimal smoothing or within-surface motion is required. The use of the PDM for segmentation across a series of subjects results in a set surfaces that retain point correspondence. The correspondence facilitates landmark-based shape analysis. Amongst other metrics, vertex-wise multivariate statistics and discriminant analysis are used to investigate local and global size and shape differences between groups. The model is fit, and shape analysis is applied to two clinical datasets.

Osteoarthritis (OA) is the most common form of human joint disease causing significant pain and disability. Current treatment for hip OA is limited to pain management and joint replacement for end-stage disease. The development of methods for early diagnosis and new treatment options are urgently needed to minimise the impact of the disease. Studies of hip OA have shown that hip joint morphology correlates with susceptibility to hip OA and disease progression. Bone shape analyses play an important role in disease diagnosis, pre-operative planning, and treatment analysis as well as in epidemiological studies aimed at identifying risk factors for hip OA. Statistical Shape Models (SSMs) are being increasingly applied to imaging-based bone shape analyses as they provide a means of quantitatively describing the global shape of the bone. This is in contrast to conventional clinical and research practice where the analysis of bone shape is reduced to a series of measurements of lengths and angles. This thesis describes the development of a novel fully automatic software system that segments the proximal femur from anteroposterior (AP) pelvic radiographs by densely placing 65 points along its contour. These annotations can then be used for the detailed morphometric analysis of proximal femur shape. The performance of the system was evaluated on a large dataset of 839 radiographs of mixed quality. Achieving a mean point-to-curve error of less than 0.9mm for 99% of all 839 AP pelvic radiographs, this is the most accurate and robust automatic method for segmenting the proximal femur in two-dimensional radiographs yet published. The system was also applied to a number of morphometric analyses of the proximal femur, showing that SSM-based radiographic proximal femur shape significantly differs between males and females, and is highly symmetric between the left and right hip joint of an individual. In addition, the research described in this thesis demonstrates how the point annotations resulting from the system can be used for univariate and multivariate genetic association analyses, identifying three novel genetic variants that contribute to radiographic proximal femur shape while also showing an association with hip OA.The developed system will facilitate complex morphometric and genetic analyses of shape variation of the proximal femur across large datasets, paving the way for the development of new options to diagnose, treat and prevent hip OA.

The thesis is focussed on the reconstruction of a damaged skull represented by a polygonal model. The reconstruction is based on a statistical shape model of the skull. The thesis covers the registration of skulls by using a thin-plate spline method, aligning polygonal models by generalized procrustes analysis, the identification of missing parts of a skull by means of statistical shape models outliers analysis. Finally, missing parts of the skull are reconstructed and the accuracy of the reconstruction is estimated.

Las enfermedades cardiovasculares (ECVs) son la principal causa de mortalidad en el mundo Occidental. El interés de prevenir y tratar las ECVs ha desencadenado un rápido desarrollo de los sistemas de adquisición de imágenes médicas. Por este motivo, la cantidad de datos de imagen recolectados en las instituciones de salud se ha incrementado considerablemente. Este hecho ha aumentado la necesidad de herramientas automatizadas para dar soporte al diagnóstico, mediante una interpretación de imagen confiable y reproducible. La tarea de interpretación requiere traducir los datos crudos de imagen en parámetros cuantitativos, los cuales son considerados relevantes para clasificar la condición cardiaca de un paciente. Para realizar tal tarea, los métodos basados en modelos estadísticos de forma han recibido favoritismo dada la naturaleza tridimensional (o 3D+t) de las imágenes cardiovasculares. Deformando el modelo estadístico de forma a la imagen de un paciente, el corazón puede analizarse de manera integral. Actualmente, el campo de las imágenes cardiovasculares esta constituido por diferentes modalidades. Cada modalidad explota diferentes fenómenos físicos, lo cual nos permite observar el órgano cardiaco desde diferentes ángulos. El personal clínico recopila todas estas piezas de información y las ensambla mentalmente en un modelo integral. Este modelo integral incluye información anatómica y funcional que muestra un cuadro completo del corazón del paciente. Es de alto interés transformar este modelo mental en un modelo computacional capaz de integrar la información de manera global. La generación de un modelo como tal no es simplemente un reto de visualización. Requiere una metodología capaz de extraer los parámetros cuantitativos relevantes basados en los mismos principios técnicos. Esto nos asegura que las mediciones se pueden comparar directamente. Tal metodología debe ser capaz de: 1) segmentar con precisión las cavidades cardiacas a partir de datos multimodales, 2) proporcionar un marco de referencia único para integrar múltiples fuentes de información, y 3) asistir la clasificación de la condición cardiaca del paciente. Esta tesis se basa en que los modelos estadísticos de forma, y en particular los Modelos Activos de Forma, son un método robusto y preciso con el potencial de incluir todos estos requerimientos. Para procesar múltiples modalidades de imagen, separamos la información estadística de forma de la información de apariencia. Obtenemos la información estadística de forma a partir de una modalidad de alta resolución y aprendemos la apariencia simulando la física de adquisición de otras modalidades. Las contribuciones de esta tesis pueden ser resumidas así: 1) un método genérico para construir automáticamente modelos de intensidad para los Modelos Activos de Forma simulando la física de adquisición de la modalidad en cuestión, 2) la primera extensión de un simulador de Resonancia Magnética Nuclear diseñado para producir estudios cardiacos realistas, y 3) un método novedoso para el entrenamiento automático de modelos de intensidad y de fiabilidad aplicado a estudios cardiacos de Resonancia Magnética Nuclear. Cada una de estas contribuciones representa un artículo publicado o enviado a una revista técnica internacional. / Cardiovascular diseases (CVDs) are the major cause of death in the Western world. The desire to prevent and treat CVDs has triggered a rapid development of medical imaging systems. As a consequence, the amount of imaging data collected in health care institutions has increased considerably. This fact has raised the need for automated analysis tools to support diagnosis with reliable and reproducible image interpretation. The interpretation task requires to translate raw imaging data into quantitative parameters, which are considered relevant to classify the patient’s cardiac condition. To achieve this task, statistical shape model approaches have found favoritism given the 3D (or 3D+t) nature of cardiovascular imaging datasets. By deforming the statistical shape model to image data from a patient, the heart can be analyzed in a more holistic way. Currently, the field of cardiovascular imaging is constituted by different modalities. Each modality exploits distinct physical phenomena, which allows us to observe the cardiac organ from different angles. Clinicians collect all these pieces of information to form an integrated mental model. The mental model includes anatomical and functional information to display a full picture of the patient’s heart. It is highly desirable to transform this mental model into a computational model able to integrate the information in a comprehensive manner. Generating such a model is not simply a visualization challenge. It requires having a methodology able to extract relevant quantitative parameters by applying the same principle. This assures that the measurements are directly comparable. Such a methodology should be able to: 1) accurately segment the cardiac cavities from multimodal datasets, 2) provide a unified frame of reference to integrate multiple information sources, and 3) aid the classification of a patient’s cardiac condition. This thesis builds upon the idea that statistical shape models, in particular Active Shape Models, are a robust and accurate approach with the potential to incorporate all these requirements. In order to handle multiple image modalities, we separate the statistical shape information from the appearance information. We obtain the statistical shape information from a high resolution modality and include the appearance information by simulating the physics of acquisition of other modalities. The contributions of this thesis can be summarized as: 1) a generic method to automatically construct intensity models for Active Shape Models based on simulating the physics of acquisition of the given imaging modality, 2) the first extension of a Magnetic Resonance Imaging (MRI) simulator tailored to produce realistic cardiac images, and 3) a novel automatic intensity model and reliability training strategy applied to cardiac MRI studies. Each of these contributions represents an article published or submitted to a peer-review archival journal.