Os métodos de controle de qualidade tradicionais aplicados ao radiodiagnóstico, é a melhor maneira de garantir a boa qualidade das imagens produzidas. No entanto, a investigação de particularidades oriundas do processo de formação de imagens radiológicas requer ferramentas computacionais complementares, em função do número de variáveis envolvidas. Entretanto, os fantomas computacionais baseados em voxels não conseguem representar as variações morfométricas necessárias para a simulação de exames cujo diagnóstico é baseado em imagem. Neste trabalho foi desenvolvido um novo tipo de fantoma computacional, baseado em modelagem 3D, que possui as vantagens apresentadas pelos fantomas computacionais tradicionais sem os problemas encontados nestes. A ferramenta de modelagem utilizada, o Blender, é disponibilizada gratuitamente na internet. A técnica utilizada foi a box modeling, que consiste na deformação de uma primitiva básica, nesse caso um cubo, até que apresente a forma da estrutura que se deseja modelar. Para tanto, foram utilizadas como referencia, imagens obtidas de atlas de anatomia e fotografias de um esqueleto fornecido pela Universidade de Mogi das Cruzes. Foram modelados o sistema ósseo, os órgãos internos e a anatomia externa do corpo humano. A metodologia empregada permitiu a alteração de parâmetros do modelo dentro da ferramenta da modelagem. Essa possibilidade foi mostrada através da variação, dos formatos do intestino e do aumento da quantidade de tecido adiposo da malha referente a pele. A simulação das imagens radiológicas foi realizada a partir de coeficientes de atenuação de massa de materiais, ossos e tecidos e de modelos com diversas características físicas. Essa versatilidade permite prever a influência que as diferenças morfométricas entre os indivíduos provocam nas imagens, propriciando dessa forma, uma ferramenta relevante complementar aos métodos de controle de qualidade tradicionais. / The conventional methods of quality control applied to radio diagnosis are the best way to have assured good quality of the produced images. Due the amount of variables to consider, the study of particular issues of the process of formation of radiological images requires complementary computational tools. However, the computational voxel based phantoms are not suitable to represent the morphometrical variations, intended for test simulations with image based diagnosis. This work developed a new type of computational phantom, based on 3D modelling. It has the same advantages of the conventional ones, without some of their restrictions. The modeling tool employed, Blender, is available on internet for free download. The project uses the technique called box modeling, which consists in the deformation of a primitive form (a cube, in this case) until it presents a similar form to that it is wanted to model. In order to achieve it, some images, obtained from anatomy atlas and a skeleton pictures obtained from University of Mogi das Cruzes, were used as reference. Were built models from skeletal system, internal organs and external human body anatomy. The applied methodology allowed model´s parameter settings on the modelling tool. This option was presented by means of intestine format variation and increase of adipose tissue on the mesh that represents skin. The simulation of radiological images was done by means of x-ray mass attenuation coefficients, bones and tissues and models with diferent physical characteristics. This flexibility allows the analysis and forecasting of the influences that morphometrical differences of individual implies on images, revealing an important tool that complements the conventional quality control tools.

abstract: Detecting anatomical structures, such as the carina, the pulmonary trunk and the aortic arch, is an important step in designing a CAD system of detection Pulmonary Embolism. The presented CAD system gets rid of the high-level prior defined knowledge to become a system which can easily extend to detect other anatomic structures. The system is based on a machine learning algorithm --- AdaBoost and a general feature --- Haar. This study emphasizes on off-line and on-line AdaBoost learning. And in on-line AdaBoost, the thesis further deals with extremely imbalanced condition. The thesis first reviews several knowledge-based detection methods, which are relied on human being's understanding of the relationship between anatomic structures. Then the thesis introduces a classic off-line AdaBoost learning. The thesis applies different cascading scheme, namely multi-exit cascading scheme. The comparison between the two methods will be provided and discussed. Both of the off-line AdaBoost methods have problems in memory usage and time consuming. Off-line AdaBoost methods need to store all the training samples and the dataset need to be set before training. The dataset cannot be enlarged dynamically. Different training dataset requires retraining the whole process. The retraining is very time consuming and even not realistic. To deal with the shortcomings of off-line learning, the study exploited on-line AdaBoost learning approach. The thesis proposed a novel pool based on-line method with Kalman filters and histogram to better represent the distribution of the samples' weight. Analysis of the performance, the stability and the computational complexity will be provided in the thesis. Furthermore, the original on-line AdaBoost performs badly in imbalanced conditions, which occur frequently in medical image processing. In image dataset, positive samples are limited and negative samples are countless. A novel Self-Adaptive Asymmetric On-line Boosting method is presented. The method utilized a new asymmetric loss criterion with self-adaptability according to the ratio of exposed positive and negative samples and it has an advanced rule to update sample's importance weight taking account of both classification result and sample's label. Compared to traditional on-line AdaBoost Learning method, the new method can achieve far more accuracy in imbalanced conditions. / Dissertation/Thesis / M.S. Computing Studies 2011

Os métodos de controle de qualidade tradicionais aplicados ao radiodiagnóstico, é a melhor maneira de garantir a boa qualidade das imagens produzidas. No entanto, a investigação de particularidades oriundas do processo de formação de imagens radiológicas requer ferramentas computacionais complementares, em função do número de variáveis envolvidas. Entretanto, os fantomas computacionais baseados em voxels não conseguem representar as variações morfométricas necessárias para a simulação de exames cujo diagnóstico é baseado em imagem. Neste trabalho foi desenvolvido um novo tipo de fantoma computacional, baseado em modelagem 3D, que possui as vantagens apresentadas pelos fantomas computacionais tradicionais sem os problemas encontados nestes. A ferramenta de modelagem utilizada, o Blender, é disponibilizada gratuitamente na internet. A técnica utilizada foi a box modeling, que consiste na deformação de uma primitiva básica, nesse caso um cubo, até que apresente a forma da estrutura que se deseja modelar. Para tanto, foram utilizadas como referencia, imagens obtidas de atlas de anatomia e fotografias de um esqueleto fornecido pela Universidade de Mogi das Cruzes. Foram modelados o sistema ósseo, os órgãos internos e a anatomia externa do corpo humano. A metodologia empregada permitiu a alteração de parâmetros do modelo dentro da ferramenta da modelagem. Essa possibilidade foi mostrada através da variação, dos formatos do intestino e do aumento da quantidade de tecido adiposo da malha referente a pele. A simulação das imagens radiológicas foi realizada a partir de coeficientes de atenuação de massa de materiais, ossos e tecidos e de modelos com diversas características físicas. Essa versatilidade permite prever a influência que as diferenças morfométricas entre os indivíduos provocam nas imagens, propriciando dessa forma, uma ferramenta relevante complementar aos métodos de controle de qualidade tradicionais. / The conventional methods of quality control applied to radio diagnosis are the best way to have assured good quality of the produced images. Due the amount of variables to consider, the study of particular issues of the process of formation of radiological images requires complementary computational tools. However, the computational voxel based phantoms are not suitable to represent the morphometrical variations, intended for test simulations with image based diagnosis. This work developed a new type of computational phantom, based on 3D modelling. It has the same advantages of the conventional ones, without some of their restrictions. The modeling tool employed, Blender, is available on internet for free download. The project uses the technique called box modeling, which consists in the deformation of a primitive form (a cube, in this case) until it presents a similar form to that it is wanted to model. In order to achieve it, some images, obtained from anatomy atlas and a skeleton pictures obtained from University of Mogi das Cruzes, were used as reference. Were built models from skeletal system, internal organs and external human body anatomy. The applied methodology allowed model´s parameter settings on the modelling tool. This option was presented by means of intestine format variation and increase of adipose tissue on the mesh that represents skin. The simulation of radiological images was done by means of x-ray mass attenuation coefficients, bones and tissues and models with diferent physical characteristics. This flexibility allows the analysis and forecasting of the influences that morphometrical differences of individual implies on images, revealing an important tool that complements the conventional quality control tools.