Lung cancer has been a major cause of death among types of cancers in the world. In the early stages, lung nodules can be detected by the aid of imaging modalities such as Computed Tomography (CT). In this stage, radiologists look for irregular rounded-shaped nodules in the lung which are normally less than 3 centimeters in diameter. Recent advancements in image analysis have proven that images contain more information than regular parameters such as intensity, histogram and morphological details. Therefore, in this project we have focused on extracting quantitative, hand-crafted features from nearly 1400 lung CT images to train a variety of classifiers based on them. In the first experiment, in total 424 Radiomics features per image has been used to train classifiers such as: Random Forest (RF), Support Vector Machine (SVM), Decision Tree (DT), Naive Bayes (NB), Linear Discriminant Analysis (LDA) and Multi-Layer Perceptron (MLP). In the second experiment, we evaluate each feature category separately with our classifiers. The third experiment includes wrapper feature selection methods (Forward/Backward/Recursive) and filter-based feature selection methods (Fisher score, Gini Index and Mutual information). They have been implemented to find the most relevant feature set in model construction. Performance of each learning method has been evaluated by accuracy score, wherewe achieved the highest accuracy of 78% with Random Forest classifier (74% in 5-fold average) and 0.82 Area Under the Receiver Operating Characteristics (AUROC) curve. After RF, NB and MLP showed the best average accuracy of 71.4% and 71% respectively.

Medical imaging is a powerful tool for clinical practice allowing in-vivo insight into a patient’s disease state. Many modalities exist, allowing for the collection of diverse information about the underlying tissue structure and/or function. Traditionally, medical professionals use visual assessment of scans to search for disease, assess relevant disease predictors and propose clinical intervention steps. However, the imaging data contain potentially useful information beyond visual assessment by trained professional. To better use the full depth of information contained in the image sets, quantitative imaging characteristics (QICs), can be extracted using mathematical and statistical operations on regions or volumes of interests. The process of using QICs is a pipeline typically involving image acquisition, segmentation, feature extraction, set qualification and analysis of informatics. These descriptors can be integrated into classification methods focused on differentiating between disease states. Lung cancer, a leading cause of death worldwide, is a clear application for advanced in-vivo imaging based classification methods. We hypothesize that QICs extracted from spatially-linked and size-standardized regions of surrounding lung tissue can improve risk assessment quality over features extracted from only the lung tumor, or nodule, regions. We require a robust and flexible pipeline for the extraction and selection of disease QICs in computed tomography (CT). This includes creating an optimized method for feature extraction, reduction, selection, and predictive analysis which could be applied to a multitude of disease imaging problems. This thesis expanded a developmental pipeline for machine learning using a large multicenter controlled CT dataset of lung nodules to extract CT QICs from the nodule, surrounding parenchyma, and greater lung volume and explore CT feature interconnectivity. Furthermore, it created a validated pipeline that is more computationally and time efficient and with stability of performance. The modularity of the optimized pipeline facilitates broader application of the tool for applications beyond CT identified pulmonary nodules. We have developed a flexible and robust pipeline for the extraction and selection of Quantitative Imaging Characteristics for Risk Assessment from the Tumor and its Environment (QIC-RATE). The results presented in this thesis support our hypothesis, showing that classification of lung and breast tumors is improved through inclusion of peritumoral signal. Optimal performance in the lung application achieved with the QIC-RATE tool incorporating 75% of the nodule diameter equivalent in perinodular parenchyma with a development performance of 100% accuracy. The stability of performance was reflected in the maintained high accuracy (98%) in the independent validation dataset of 100 CT from a separate institution. In the breast QIC-RATE application, optimal performance was achieved using 25% of the tumor diameter in breast tissue with 90% accuracy in development, 82% in validation. We address the need for more complex assessments of medically imaged tumors through the QIC-RATE pipeline; a modular, scalable, transferrable pipeline for extracting, reducing and selecting, and training a classification tool based on QICs. Altogether, this research has resulted in a risk assessment methodology that is validated, stable, high performing, adaptable, and transparent.

Purpose. Low-dose computed tomography is an important imaging modality for screening and surveillance of lung cancer. The goal of this study was to determine the extent to which dose could be minimized while maintaining diagnostic accuracy through knowledgeable selection of reconstruction techniques. Methods. An anthropomorphic phantom was imaged on a 320-slice volumetric CT scanner. Detectability of small solid lung nodules was evaluated as a function of dose, patient size, reconstruction filter and slice thickness by means of 9-alternative forced-choice observer tests. Results. Nodule detectability decreased sharply below a threshold dose level due to increased image noise. For large body habitus, optimal (smooth) filter selection reduced dose by a factor of ~3. Nodule detectability decreased for slice thicknesses larger than the nodule diameter. Conclusions. Radiation dose can be reduced well below current clinical protocols. Smooth reconstruction filters and avoidance of large slice thickness permits lower-dose techniques without tradeoff in diagnostic performance.

Purpose. Low-dose computed tomography is an important imaging modality for screening and surveillance of lung cancer. The goal of this study was to determine the extent to which dose could be minimized while maintaining diagnostic accuracy through knowledgeable selection of reconstruction techniques. Methods. An anthropomorphic phantom was imaged on a 320-slice volumetric CT scanner. Detectability of small solid lung nodules was evaluated as a function of dose, patient size, reconstruction filter and slice thickness by means of 9-alternative forced-choice observer tests. Results. Nodule detectability decreased sharply below a threshold dose level due to increased image noise. For large body habitus, optimal (smooth) filter selection reduced dose by a factor of ~3. Nodule detectability decreased for slice thicknesses larger than the nodule diameter. Conclusions. Radiation dose can be reduced well below current clinical protocols. Smooth reconstruction filters and avoidance of large slice thickness permits lower-dose techniques without tradeoff in diagnostic performance.

京都大学 / 0048 / 新制・課程博士 / 博士(医学) / 甲第18155号 / 医博第3875号 / 新制||医||1002(附属図書館) / 31013 / 京都大学大学院医学研究科医学専攻 / (主査)教授 上本 伸二, 教授 平岡 眞寛, 教授 安達 泰治 / 学位規則第4条第1項該当 / Doctor of Medical Science / Kyoto University / DFAM

No description available.

Submitted by Maria Aparecida (cidazen@gmail.com) on 2017-08-18T14:02:37Z No. of bitstreams: 1 cleriston.pdf: 1605933 bytes, checksum: c1faa5f854c1a9debfbaa1affc5ab4ad (MD5) / Made available in DSpace on 2017-08-18T14:02:37Z (GMT). No. of bitstreams: 1 cleriston.pdf: 1605933 bytes, checksum: c1faa5f854c1a9debfbaa1affc5ab4ad (MD5) Previous issue date: 2009-02-12 / The diagnosis of lung nodules has been constantly looked for by researchers as a way to minimize the high global mortality indices related to lung cancer. The usage of medical images, such as Computerized Tomography, has made possible the deepening and the improvement of techniques used to evaluate exams and provide diagnosis. This work presents a methodology for diagnosing single lung nodules that can be an aid for studies performed on similar areas and for specialists. This methodology was applied to two different image databases. The representation of the nodules was done with extraction of geometry and texture features, being the last one described through Simpson’s Index, a statistic used in Spatial Analysis and in Ecology. These features were submitted to the Support Vector Machine classifier (SVM) in two approaches: the traditional approach and the approach by using One Class. With the traditional SVM approach, we have obtained sensibility rates of 90%, specificity of 96.67% and accuracy of 95%. Using One Class SVM, the obtained rates were: sensibility of 89.7%, specificity of 89.7% and accuracy of 89.7%. / O diagnóstico de nódulos pulmonares tem sido buscado constantemente por pesquisadores como forma de amenizar os altos índices de mortalidade mundial relacionado ao câncer de pulmão. O uso de imagens médicas, como a Tomografia Computadorizada, tem possibilitado um aprofundamento e melhoramento de técnicas para avaliar exames e prover diagnósticos. Este trabalho apresenta uma metodologia para diagnóstico de nódulos pulmonares solitários que possa servir como um auxílio para estudos realizados em áreas afins e para especialistas. Esta metodologia foi aplicada a duas diferentes bases de dados de imagens. A representação dos nódulos foi feita com a extração de medidas de geometria e de textura sendo esta última descrita através do Índice de Simpson, uma estatística utilizada na Análise Espacial e na Ecologia. Essas medidas foram submetidas ao classificador Máquina de Vetores de Suporte - MVS em duas abordagens: a abordagem tradicional e abordagem usando uma classe. Com abordagem MVS tradicional, obtiveramse taxas de sensibilidade de 90%, especificidade de 96,67% e acurácia de 95%. Usando MVS de uma classe, as taxas obtidas foram: sensibilidade igual a 89,7%, especificidade igual a 89,7% e acurácia igual a 89,7%.

Made available in DSpace on 2016-08-17T14:53:18Z (GMT). No. of bitstreams: 1 dissertacao Alex Martins Santos.pdf: 5806595 bytes, checksum: 431584590b26aa21a56d684cd9c7f282 (MD5) Previous issue date: 2011-08-19 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Lung cancer stands out by pointing the highest incidence and higher mortality rate of all other types of cancer. It has one of the lower survival rates of after diagnosis, which is mainly due to late detection and therefore delayed treatment. Computer-aided detection systems (CAD) are developed to assist the specialist in the search and identification of nodules and changes in CT. These systems respectively aim to automate the identification and classification of these structures. This work aims to study and develop a methodology for automatic detection of small lung nodules (bigger than 2 mm and smaller than 10 mm in diameter). The proposed methodology is based on techniques of image processing and pattern recognition. Similar applications use widely some of these techniques. The proposed methodology also uses other techniques from different areas and applications, such as measures of the Tsallis and Shannon entropy used in this study to describe suspected structures. These measures are respectively provided from statistical mechanics and information theory, however lately they have been successfully applied in image processing. It was also used the Gaussian mixture model (GMM) and the Hessian matrix calculation to separate the internal structures of the remaining lung parenchyma. Promising results were found in tests with 140 exams divided in of 80% for training and 20% for testing. It was achieved a 79% of sensitivity rate and a total of 1.17 false positives per slice. / Dentre os outros tipos de câncer, o câncer de pulmão se destaca por apresentar a maior incidência e a maior taxa de mortalidade de todos, além de uma das menores taxas de sobrevida após o diagnóstico (cinco anos em média), fato este decorrido principalmente pela detecção e, conseqüentemente, tratamento tardio. Para auxiliar o especialista na busca e identificação de nódulos e alterações em imagens tomográficas são desenvolvidos sistemas de detecção auxiliados por computador (CAD) que visam automatizar os trabalhos de identificação e classificação de dessas estruturas. O presente trabalho tem por objetivo o estudo e desenvolvimento de uma metodologia para detecção automática de nódulos pulmonares pequenos (maiores que dois milímetros e menores que 10 milímetros de diâmetro). A metodologia proposta se baseia em técnicas de processamento de imagens e reconhecimento de padrões. Algumas dessas técnicas são amplamente utilizadas em aplicações similares, já outras técnicas utilizadas provêm de outras áreas e aplicações, como é o caso das medidas de entropia de Tsallis e Shannon, utilizados neste trabalho para descrever estruturas suspeitas. Estas medidas provém respectivamente da mecânica estatística e da Teoria da Informação, porém ultimamente tem sido aplicadas com sucesso no processamento de imagens. Também foi empregado o Modelo de Misturas Gaussianas (GMM) e o cálculo da matriz Hessiana para separar as estruturas internas do pulmão do restante do parênquima. Resultados promissores foram encontrados, em teste com 140 exames divididos em 80% para treino e 20% para testes, alcançou-se uma sensibilidade de 79% e um total de 1,17 falsos positivos por fatia.

Made available in DSpace on 2016-08-17T14:53:26Z (GMT). No. of bitstreams: 1 dissertacao Pedro Henique.pdf: 3622658 bytes, checksum: 95066b31ffa29d90dfade445121817be (MD5) Previous issue date: 2014-01-03 / With the advent of imaging methods, the pulmonary nodule is becoming the most common manifestation of lung cancer and one of the most lethal of all cancers. Geometry (shape) and texture (tissue) measurements analyzed over time can be used to search the nodule malignancy. Among geometric measures commonly used, the nodule growth rate is one of the most accurate noninvasive methods to evaluate malignancy. Followed by other texture measures achieved over time, it is possible to get valuable information about nodules behavior, so that the doctor can use them to take related decisions. For these reasons, it is important to compare the nodule in exams applied at different moments. A key step for the comparison is to verify the correspondence between nodules of different exams. This correspondence is used to determine if a nodule X in the exam A is the same nodule Y in an exam B. Due to a number of anatomical and physiological factors and image acquisition, the same nodule cannot be in exactly the same location on different exams. To correct this problem, rigid and deformable image registration show up to be efficient. Once established this correspondence, it is possible to analyze the nodule texture changes through similarity indexes. In this sense, the aim of this work is to present methods for quantitative analysis of texture changes in lung nodules. For this analysis, it is used CT scans obtained at different moments from the same patient. Furthermore, it is presented a method to verify nodules found in different exams correspond to the same nodule by applying image registration. / Com o advento dos métodos de imagem, o nódulo pulmonar vem se tornando a manifestação mais comum de câncer de pulmão e um dos mais letais de todos os cânceres. Uma forma de pesquisar a malignidade de um nódulo é analisar temporalmente suas medidas de geometria (forma) e textura (tecido). Entre medidas geométricas comumente utilizadas, a taxa de crescimento do nódulo constitui um dos métodos mais precisos não invasivos de aferição da malignidade. Acompanhada a outras medidas de textura obtidas no decorrer do tempo, obtém-se informações valiosas sobre o seu comportamento, de forma que o médico pode usar essas medidas na tomada de decisões. Sabendo disso, é importante a comparação do nódulo em exames extraídos em momentos diferentes. Uma etapa fundamental para essa comparação é verificar a correspondência entre nódulos em exames diferentes, de forma que seja possível determinar se um nódulo X em um exame A é o mesmo nódulo Y em um exame B. Devido a uma série de fatores anátomo-fisiológicos e de aquisição de imagens, um mesmo nódulo pode não estar exatamente na mesma localização em exames diferentes. Para corrigir esse problema, registros de imagem rígidos e deformáveis mostram-se eficientes. Uma vez estabelecida essa correspondência, é possível analisar as mudanças na textura do nódulo através de índices de similaridade. Nesse sentido, o objetivo desse trabalho é apresentar métodos para a análise quantitativa de mudanças de textura em nódulos. Para essa análise serão utilizadas imagens de tomografia computadorizada obtidas em momentos diferentes de um mesmo paciente. Além disso, será apresentado um método para verificar se nódulos encontrados em exames diferentes correspondem ao mesmo nódulo através da aplicação de registros de imagens.

Made available in DSpace on 2016-08-17T14:53:29Z (GMT). No. of bitstreams: 1 Joao Rodrigo Ferreira.pdf: 1547613 bytes, checksum: a9c7e73154b7e9a72733f9f2e20a55fe (MD5) Previous issue date: 2007-12-07 / The lung cancer is a disorder with significant prevalence in several countries worldwide. The hard treatment and the fast progress of the disease increase the mortality rates. The main factor contributing to a successful treatment is an early diagnosis. However possible omissions in the scan analysis can lead to late diagnosis, compromising all the treatment. In order to present a computational tool aimed at nodules detection, that can be used as a second opinion to the specialist, this master thesis proposes a methodology for nodules detection that is totally automatic, robust and consistent. The methodology is based on successive refinements for the segmentation of computed tomography images using morphologic techniques to obtain nodule candidates. The false positive reduction is achieved by SVM based on geometric and texture features. The tests, performed with real scans, indicate the feasibility of the proposed method. In automatic detection performed on 33 cases the methodology reached 95.21% of correctness with 0.42 false positives and 0.15 false negative per scan. / O câncer de pulmão é uma enfermidade com prevalência significativa em diversos países no mundo todo. O difícil tratamento e a progressão rápida da doença fazem com que os índices de mortalidade das pessoas acometidas por este mal sejam muito altos. O principal fator contribuinte para um tratamento de sucesso, entretanto, é o diagnóstico precoce. Contudo possíveis omissões na análise dos exames podem levar a um diagnóstico tardio, comprometendo todo o tratamento. Com o intuito de oferecer uma alternativa computacional de auxílio à detecção de nódulos, servindo como uma segunda opinião para o médico, este trabalho propõe uma metodologia totalmente automática, robusta e consistente. A metodologia é fundamentada em refinamentos sucessivos da segmentação sobre imagens de tomografia computadorizada utilizando técnicas morfológicas para a obtenção de candidatos a nódulo. A redução de falsos positivos é efetivada pelo SVM com base em características geométricas e de textura. Os testes realizados com exames reais indicam a viabilidade da solução proposta. Na detecção automática realizada sobre 33 casos a metodologia atingiu 95,21% de acerto com uma média de 0,42 falsos positivos e 0,15 falsos negativos por exame.