Automatic segmentation of abdominal aortic aneurysm from medical images based on active shape models and texture models

A semi-automatic segmentation algorithm for abdominal aortic aneurysms (AAA), and based on Active Shape Models (ASM) and texture models, is presented in this work. The texture information is provided by a set of four 3D magnetic resonance (MR) images, composed of axial slices of the abdomen, where lumen, wall and intraluminal thrombus (ILT) are visible. Due to the reduced number of images in the MRI training set, an ASM and a custom texture model based on border intensity statistics are constructed. For the same reason the shape is characterized from 35-computed tomography angiography (CTA) images set so the shape variations are better represented. For the evaluation, leave-one-out experiments have been held over the four MRI set.

A method for incorporating residual stresses into patient-specific finite element simulations of arteries with an example on AAAs

Through progress in medical imaging, image analysis and finite element (FE) meshing tools it is now possible to extract patient-specific geometries from medical images of abdominal aortic aneurysms(AAAs), and thus to study clinically-relevant problems via FE simulations. Such simulations allow additional insight into human physiology in both healthy and diseased states. Medical imaging is most often performed in vivo, and hence the reconstructed model geometry in the problem of interest will represent the in vivo state, e.g., the AAA at physiological blood pressure. However, classical continuum mechanics and FE methods assume that constitutive models and the corresponding simulations begin from an unloaded, stress-free reference condition.

Optimización de una energía mediante cortes de grafos. Segmentación de imágenes

La segmentación de imágenes puede plantearse como un problema de minimización de una energía discreta. Nos enfrentamos así a una doble cuestión: definir una energía cuyo mínimo proporcione la segmentación buscada y, una vez definida la energía, encontrar un mínimo absoluto de la misma. La primera parte de esta tesis aborda el segundo problema, y la segunda parte, en un contexto más aplicado, el primero. Las técnicas de minimización basadas en cortes de grafos permiten obtener el mínimo de una energía discreta en tiempo polinomial mediante algoritmos de tipo min-cut/max-flow. Sin embargo, estas técnicas solo pueden aplicarse a energías que son representabas por grafos. Un importante reto es estudiar qué energías son representabas así como encontrar un grafo que las represente, lo que equivale a encontrar una función gadget con variables adicionales. En la primera parte de este trabajo se estudian propiedades de las funciones gadgets que permiten acotar superiormente el número de variables adicionales. Además se caracterizan las energías con cuatro variables que son representabas, definiendo gadgets con dos variables adicionales. En la segunda parte, más práctica, se aborda el problema de segmentación de imágenes médicas, base en muchas ocasiones para la diagnosis y el seguimiento de terapias. La segmentación multi-atlas es una potente técnica de segmentación automática de imágenes médicas, con tres aspectos importantes a destacar: el tipo de registro entre los atlas y la imagen objetivo, la selección de atlas y el método de fusión de etiquetas. Este último punto puede formularse como un problema de minimización de una energía. A este respecto introducimos dos nuevas energías representables. La primera, de orden dos, se utiliza en la segmentación en hígado y fondo de imágenes abdominales obtenidas mediante tomografía axial computarizada. La segunda, de orden superior, se utiliza en la segmentación en hipocampos y fondo de imágenes cerebrales obtenidas mediante resonancia magnética. ABSTRACT The image segmentation can be described as the problem of minimizing a discrete energy. We face two problems: first, to define an energy whose minimum provides the desired segmentation and, second, once the energy is defined we must find its global minimum. The first part of this thesis addresses the second problem, and the second part, in a more applied context, the first problem. Minimization techniques based on graph cuts find the minimum of a discrete energy in polynomial time via min-cut/max-flow algorithms. Nevertheless, these techniques can only be applied to graph-representable energies. An important challenge is to study which energies are graph-representable and to construct graphs which represent these energies. This is the same as finding a gadget function with additional variables. In the first part there are studied the properties of gadget functions which allow the number of additional variables to be bounded from above. Moreover, the graph-representable energies with four variables are characterised and gadgets with two additional variables are defined for these. The second part addresses the application of these ideas to medical image segmentation. This is often the first step in computer-assisted diagnosis and monitoring therapy. Multiatlas segmentation is a powerful automatic segmentation technique for medical images, with three important aspects that are highlighted here: the registration between the atlas and the target image, the atlas selection, and the label fusion method. We formulate the label fusion method as a minimization problem and we introduce two new graph-representable energies. The first is a second order energy and it is used for the segmentation of the liver in computed tomography (CT) images. The second energy is a higher order energy and it is used for the segmentation of the hippocampus in magnetic resonance images (MRI).

Procesado de imágenes médicas en MATLAB

La relación entre la ingeniería y la medicina cada vez se está haciendo más estrecha, y debido a esto se ha creado una nueva disciplina, la bioingeniería, ámbito en el que se centra el proyecto. Este ámbito cobra gran interés debido al rápido desarrollo de nuevas tecnologías que en particular permiten, facilitan y mejoran la obtención de diagnósticos médicos respecto de los métodos tradicionales. Dentro de la bioingeniería, el campo que está teniendo mayor desarrollo es el de la imagen médica, gracias al cual se pueden obtener imágenes del interior del cuerpo humano con métodos no invasivos y sin necesidad de recurrir a la cirugía. Mediante métodos como la resonancia magnética, rayos X, medicina nuclear o ultrasonidos, se pueden obtener imágenes del cuerpo humano para realizar diagnósticos. Para que esas imágenes puedan ser utilizadas con ese fin hay que realizar un correcto tratamiento de éstas mediante técnicas de procesado digital. En ése ámbito del procesado digital de las imágenes médicas es en el que se ha realizado este proyecto. Gracias al desarrollo del tratamiento digital de imágenes con métodos de extracción de información, mejora de la visualización o resaltado de rasgos de interés de las imágenes, se puede facilitar y mejorar el diagnóstico de los especialistas. Por todo esto en una época en la que se quieren automatizar todos los procesos para mejorar la eficacia del trabajo realizado, el automatizar el procesado de las imágenes para extraer información con mayor facilidad, es muy útil. Actualmente una de las herramientas más potentes en el tratamiento de imágenes médicas es Matlab, gracias a su toolbox de procesado de imágenes. Por ello se eligió este software para el desarrollo de la parte práctica de este proyecto, su potencia y versatilidad simplifican la implementación de algoritmos. Este proyecto se estructura en dos partes. En la primera se realiza una descripción general de las diferentes modalidades de obtención de imágenes médicas y se explican los diferentes usos de cada método, dependiendo del campo de aplicación. Posteriormente se hace una descripción de las técnicas más importantes de procesado de imagen digital que han sido utilizadas en el proyecto. En la segunda parte se desarrollan cuatro aplicaciones en Matlab para ejemplificar el desarrollo de algoritmos de procesado de imágenes médicas. Dichas implementaciones demuestran la aplicación y utilidad de los conceptos explicados anteriormente en la parte teórica, como la segmentación y operaciones de filtrado espacial de la imagen, así como otros conceptos específicos. Las aplicaciones ejemplo desarrolladas han sido: obtención del porcentaje de metástasis de un tejido, diagnóstico de las deformidades de la columna vertebral, obtención de la MTF de una cámara de rayos gamma y medida del área de un fibroadenoma de una ecografía de mama. Por último, para cada una de las aplicaciones se detallará su utilidad en el campo de la imagen médica, los resultados obtenidos y su implementación en una interfaz gráfica para facilitar su uso. ABSTRACT. The relationship between medicine and engineering is becoming closer than ever giving birth to a recently appeared science field: bioengineering. This project is focused on this subject. This recent field is becoming more and more important due to the fast development of new technologies that provide tools to improve disease diagnosis, with regard to traditional procedures. In bioengineering the fastest growing field is medical imaging, in which we can obtain images of the inside of the human body without need of surgery. Nowadays by means of the medical modalities of magnetic resonance, X ray, nuclear medicine or ultrasound, we can obtain images to make a more accurate diagnosis. For those images to be useful within the medical field, they should be processed properly with some digital image processing techniques. It is in this field of digital medical image processing where this project is developed. Thanks to the development of digital image processing providing methods for data collection, improved visualization or data highlighting, diagnosis can be eased and facilitated. In an age where automation of processes is much sought, automated digital image processing to ease data collection is extremely useful. One of the most powerful image processing tools is Matlab, together with its image processing toolbox. That is the reason why that software was chosen to develop the practical algorithms in this project. This final project is divided into two main parts. Firstly, the different modalities for obtaining medical images will be described. The different usages of each method according to the application will also be specified. Afterwards we will give a brief description of the most important image processing tools that have been used in the project. Secondly, four algorithms in Matlab are implemented, to provide practical examples of medical image processing algorithms. This implementation shows the usefulness of the concepts previously explained in the first part, such as: segmentation or spatial filtering. The particular applications examples that have been developed are: calculation of the metastasis percentage of a tissue, diagnosis of spinal deformity, approximation to the MTF of a gamma camera, and measurement of the area of a fibroadenoma in an ultrasound image. Finally, for each of the applications developed, we will detail its usefulness within the medical field, the results obtained, and its implementation in a graphical user interface to ensure ease of use.

High-quality conforming hexahedral meshes of patient-specific abdominal aortic aneurysms including their intraluminal thrombi

In order to perform finite element (FE) analyses of patient-specific abdominal aortic aneurysms, geometries derived from medical images must be meshed with suitable elements. We propose a semi-automatic method for generating conforming hexahedral meshes directly from contours segmented from medical images. Magnetic resonance images are generated using a protocol developed to give the abdominal aorta high contrast against the surrounding soft tissue. These data allow us to distinguish between the different structures of interest. We build novel quadrilateral meshes for each surface of the sectioned geometry and generate conforming hexahedral meshes by combining the quadrilateral meshes. The three-layered morphology of both the arterial wall and thrombus is incorporated using parameters determined from experiments. We demonstrate the quality of our patient-specific meshes using the element Scaled Jacobian. The method efficiently generates high-quality elements suitable for FE analysis, even in the bifurcation region of the aorta into the iliac arteries. For example, hexahedral meshes of up to 125,000 elements are generated in less than 130 s, with 94.8 % of elements well suited for FE analysis. We provide novel input for simulations by independently meshing both the arterial wall and intraluminal thrombus of the aneurysm, and their respective layered morphologies.

Cloud Computing Service for Managing Large Medical Image Data-Sets Using Balanced Collaborative Agents

Managing large medical image collections is an increasingly demanding important issue in many hospitals and other medical settings. A huge amount of this information is daily generated, which requires robust and agile systems. In this paper we present a distributed multi-agent system capable of managing very large medical image datasets. In this approach, agents extract low-level information from images and store them in a data structure implemented in a relational database. The data structure can also store semantic information related to images and particular regions. A distinctive aspect of our work is that a single image can be divided so that the resultant sub-images can be stored and managed separately by different agents to improve performance in data accessing and processing. The system also offers the possibility of applying some region-based operations and filters on images, facilitating image classification. These operations can be performed directly on data structures in the database.

DAEDALUS at ImageCLEF Medical Retrieval 2011: Textual, Visual and Multimodal Experiments

This paper describes the participation of DAEDALUS at ImageCLEF 2011 Medical Retrieval task. We have focused on multimodal (or mixed) experiments that combine textual and visual retrieval. The main objective of our research has been to evaluate the effect on the medical retrieval process of the existence of an extended corpus that is annotated with the image type, associated to both the image itself and also to its textual description. For this purpose, an image classifier has been developed to tag each document with its class (1st level of the hierarchy: Radiology, Microscopy, Photograph, Graphic, Other) and subclass (2nd level: AN, CT, MR, etc.). For the textual-based experiments, several runs using different semantic expansion techniques have been performed. For the visual-based retrieval, different runs are defined by the corpus used in the retrieval process and the strategy for obtaining the class and/or subclass. The best results are achieved in runs that make use of the image subclass based on the classification of the sample images. Although different multimodal strategies have been submitted, none of them has shown to be able to provide results that are at least comparable to the ones achieved by the textual retrieval alone. We believe that we have been unable to find a metric for the assessment of the relevance of the results provided by the visual and textual processes

Crowdsourcing malaria parasite quantification: an online game for analyzing images of infected thick blood smears

Background: There are 600,000 new malaria cases daily worldwide. The gold standard for estimating the parasite burden and the corresponding severity of the disease consists in manually counting the number of parasites in blood smears through a microscope, a process that can take more than 20 minutes of an expert microscopist’s time. Objective: This research tests the feasibility of a crowdsourced approach to malaria image analysis. In particular, we investigated whether anonymous volunteers with no prior experience would be able to count malaria parasites in digitized images of thick blood smears by playing a Web-based game. Methods: The experimental system consisted of a Web-based game where online volunteers were tasked with detecting parasites in digitized blood sample images coupled with a decision algorithm that combined the analyses from several players to produce an improved collective detection outcome. Data were collected through the MalariaSpot website. Random images of thick blood films containing Plasmodium falciparum at medium to low parasitemias, acquired by conventional optical microscopy, were presented to players. In the game, players had to find and tag as many parasites as possible in 1 minute. In the event that players found all the parasites present in the image, they were presented with a new image. In order to combine the choices of different players into a single crowd decision, we implemented an image processing pipeline and a quorum algorithm that judged a parasite tagged when a group of players agreed on its position. Results: Over 1 month, anonymous players from 95 countries played more than 12,000 games and generated a database of more than 270,000 clicks on the test images. Results revealed that combining 22 games from nonexpert players achieved a parasite counting accuracy higher than 99%. This performance could be obtained also by combining 13 games from players trained for 1 minute. Exhaustive computations measured the parasite counting accuracy for all players as a function of the number of games considered and the experience of the players. In addition, we propose a mathematical equation that accurately models the collective parasite counting performance. Conclusions: This research validates the online gaming approach for crowdsourced counting of malaria parasites in images of thick blood films. The findings support the conclusion that nonexperts are able to rapidly learn how to identify the typical features of malaria parasites in digitized thick blood samples and that combining the analyses of several users provides similar parasite counting accuracy rates as those of expert microscopists. This experiment illustrates the potential of the crowdsourced gaming approach for performing routine malaria parasite quantification, and more generally for solving biomedical image analysis problems, with future potential for telediagnosis related to global health challenges.

MIA - a free and open source software for gray scale medical image analysis

Background Gray scale images make the bulk of data in bio-medical image analysis, and hence, the main focus of many image processing tasks lies in the processing of these monochrome images. With ever improving acquisition devices, spatial and temporal image resolution increases, and data sets become very large. Various image processing frameworks exists that make the development of new algorithms easy by using high level programming languages or visual programming. These frameworks are also accessable to researchers that have no background or little in software development because they take care of otherwise complex tasks. Specifically, the management of working memory is taken care of automatically, usually at the price of requiring more it. As a result, processing large data sets with these tools becomes increasingly difficult on work station class computers. One alternative to using these high level processing tools is the development of new algorithms in a languages like C++, that gives the developer full control over how memory is handled, but the resulting workflow for the prototyping of new algorithms is rather time intensive, and also not appropriate for a researcher with little or no knowledge in software development. Another alternative is in using command line tools that run image processing tasks, use the hard disk to store intermediate results, and provide automation by using shell scripts. Although not as convenient as, e.g. visual programming, this approach is still accessable to researchers without a background in computer science. However, only few tools exist that provide this kind of processing interface, they are usually quite task specific, and don’t provide an clear approach when one wants to shape a new command line tool from a prototype shell script. Results The proposed framework, MIA, provides a combination of command line tools, plug-ins, and libraries that make it possible to run image processing tasks interactively in a command shell and to prototype by using the according shell scripting language. Since the hard disk becomes the temporal storage memory management is usually a non-issue in the prototyping phase. By using string-based descriptions for filters, optimizers, and the likes, the transition from shell scripts to full fledged programs implemented in C++ is also made easy. In addition, its design based on atomic plug-ins and single tasks command line tools makes it easy to extend MIA, usually without the requirement to touch or recompile existing code. Conclusion In this article, we describe the general design of MIA, a general purpouse framework for gray scale image processing. We demonstrated the applicability of the software with example applications from three different research scenarios, namely motion compensation in myocardial perfusion imaging, the processing of high resolution image data that arises in virtual anthropology, and retrospective analysis of treatment outcome in orthognathic surgery. With MIA prototyping algorithms by using shell scripts that combine small, single-task command line tools is a viable alternative to the use of high level languages, an approach that is especially useful when large data sets need to be processed.

Pulmonary Artery-Vein Segmentation in Real and Synthetic CT Images = Segmentación Arteria-Vena Pulmonares en Imágenes de CT Reales y Sintéticas

La tomografía axial computerizada (TAC) es la modalidad de imagen médica preferente para el estudio de enfermedades pulmonares y el análisis de su vasculatura. La segmentación general de vasos en pulmón ha sido abordada en profundidad a lo largo de los últimos años por la comunidad científica que trabaja en el campo de procesamiento de imagen; sin embargo, la diferenciación entre irrigaciones arterial y venosa es aún un problema abierto. De hecho, la separación automática de arterias y venas está considerado como uno de los grandes retos futuros del procesamiento de imágenes biomédicas. La segmentación arteria-vena (AV) permitiría el estudio de ambas irrigaciones por separado, lo cual tendría importantes consecuencias en diferentes escenarios médicos y múltiples enfermedades pulmonares o estados patológicos. Características como la densidad, geometría, topología y tamaño de los vasos sanguíneos podrían ser analizados en enfermedades que conllevan remodelación de la vasculatura pulmonar, haciendo incluso posible el descubrimiento de nuevos biomarcadores específicos que aún hoy en dípermanecen ocultos. Esta diferenciación entre arterias y venas también podría ayudar a la mejora y el desarrollo de métodos de procesamiento de las distintas estructuras pulmonares. Sin embargo, el estudio del efecto de las enfermedades en los árboles arterial y venoso ha sido inviable hasta ahora a pesar de su indudable utilidad. La extrema complejidad de los árboles vasculares del pulmón hace inabordable una separación manual de ambas estructuras en un tiempo realista, fomentando aún más la necesidad de diseñar herramientas automáticas o semiautomáticas para tal objetivo. Pero la ausencia de casos correctamente segmentados y etiquetados conlleva múltiples limitaciones en el desarrollo de sistemas de separación AV, en los cuales son necesarias imágenes de referencia tanto para entrenar como para validar los algoritmos. Por ello, el diseño de imágenes sintéticas de TAC pulmonar podría superar estas dificultades ofreciendo la posibilidad de acceso a una base de datos de casos pseudoreales bajo un entorno restringido y controlado donde cada parte de la imagen (incluyendo arterias y venas) está unívocamente diferenciada. En esta Tesis Doctoral abordamos ambos problemas, los cuales están fuertemente interrelacionados. Primero se describe el diseño de una estrategia para generar, automáticamente, fantomas computacionales de TAC de pulmón en humanos. Partiendo de conocimientos a priori, tanto biológicos como de características de imagen de CT, acerca de la topología y relación entre las distintas estructuras pulmonares, el sistema desarrollado es capaz de generar vías aéreas, arterias y venas pulmonares sintéticas usando métodos de crecimiento iterativo, que posteriormente se unen para formar un pulmón simulado con características realistas. Estos casos sintéticos, junto a imágenes reales de TAC sin contraste, han sido usados en el desarrollo de un método completamente automático de segmentación/separación AV. La estrategia comprende una primera extracción genérica de vasos pulmonares usando partículas espacio-escala, y una posterior clasificación AV de tales partículas mediante el uso de Graph-Cuts (GC) basados en la similitud con arteria o vena (obtenida con algoritmos de aprendizaje automático) y la inclusión de información de conectividad entre partículas. La validación de los fantomas pulmonares se ha llevado a cabo mediante inspección visual y medidas cuantitativas relacionadas con las distribuciones de intensidad, dispersión de estructuras y relación entre arterias y vías aéreas, los cuales muestran una buena correspondencia entre los pulmones reales y los generados sintéticamente. La evaluación del algoritmo de segmentación AV está basada en distintas estrategias de comprobación de la exactitud en la clasificación de vasos, las cuales revelan una adecuada diferenciación entre arterias y venas tanto en los casos reales como en los sintéticos, abriendo así un amplio abanico de posibilidades en el estudio clínico de enfermedades cardiopulmonares y en el desarrollo de metodologías y nuevos algoritmos para el análisis de imágenes pulmonares. ABSTRACT Computed tomography (CT) is the reference image modality for the study of lung diseases and pulmonary vasculature. Lung vessel segmentation has been widely explored by the biomedical image processing community, however, differentiation of arterial from venous irrigations is still an open problem. Indeed, automatic separation of arterial and venous trees has been considered during last years as one of the main future challenges in the field. Artery-Vein (AV) segmentation would be useful in different medical scenarios and multiple pulmonary diseases or pathological states, allowing the study of arterial and venous irrigations separately. Features such as density, geometry, topology and size of vessels could be analyzed in diseases that imply vasculature remodeling, making even possible the discovery of new specific biomarkers that remain hidden nowadays. Differentiation between arteries and veins could also enhance or improve methods processing pulmonary structures. Nevertheless, AV segmentation has been unfeasible until now in clinical routine despite its objective usefulness. The huge complexity of pulmonary vascular trees makes a manual segmentation of both structures unfeasible in realistic time, encouraging the design of automatic or semiautomatic tools to perform the task. However, this lack of proper labeled cases seriously limits in the development of AV segmentation systems, where reference standards are necessary in both algorithm training and validation stages. For that reason, the design of synthetic CT images of the lung could overcome these difficulties by providing a database of pseudorealistic cases in a constrained and controlled scenario where each part of the image (including arteries and veins) is differentiated unequivocally. In this Ph.D. Thesis we address both interrelated problems. First, the design of a complete framework to automatically generate computational CT phantoms of the human lung is described. Starting from biological and imagebased knowledge about the topology and relationships between structures, the system is able to generate synthetic pulmonary arteries, veins, and airways using iterative growth methods that can be merged into a final simulated lung with realistic features. These synthetic cases, together with labeled real CT datasets, have been used as reference for the development of a fully automatic pulmonary AV segmentation/separation method. The approach comprises a vessel extraction stage using scale-space particles and their posterior artery-vein classification using Graph-Cuts (GC) based on arterial/venous similarity scores obtained with a Machine Learning (ML) pre-classification step and particle connectivity information. Validation of pulmonary phantoms from visual examination and quantitative measurements of intensity distributions, dispersion of structures and relationships between pulmonary air and blood flow systems, show good correspondence between real and synthetic lungs. The evaluation of the Artery-Vein (AV) segmentation algorithm, based on different strategies to assess the accuracy of vessel particles classification, reveal accurate differentiation between arteries and vein in both real and synthetic cases that open a huge range of possibilities in the clinical study of cardiopulmonary diseases and the development of methodological approaches for the analysis of pulmonary images.