Submitted by Erika Demachki (erikademachki@gmail.com) on 2014-10-14T21:24:19Z No. of bitstreams: 2 Dissertação - Rafael Teixeira Sousa - 2013.pdf: 2536972 bytes, checksum: 5a0aa0899207e8f66f11c5b819fcc211 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) / Approved for entry into archive by Jaqueline Silva (jtas29@gmail.com) on 2014-10-16T18:20:52Z (GMT) No. of bitstreams: 2 Dissertação - Rafael Teixeira Sousa - 2013.pdf: 2536972 bytes, checksum: 5a0aa0899207e8f66f11c5b819fcc211 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) / Made available in DSpace on 2014-10-16T18:20:52Z (GMT). No. of bitstreams: 2 Dissertação - Rafael Teixeira Sousa - 2013.pdf: 2536972 bytes, checksum: 5a0aa0899207e8f66f11c5b819fcc211 (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) Previous issue date: 2013-09-20 / This work extends a Computer-Aided Diagnosis system called PneumoCAD for detecting pneumonia in infants using radiographic images, with the aim of improving the system’s accuracy, robustness and test the features previously extracted. We implement and compare five contemporary machine learning classifiers, namely: Naïve Bayes, K-Nearest Neighbor (KNN), Support Vector Machines (SVM), Multi-Layer Perceptron (MLP) and Decision Tree, combined with three dimensionality reduction algorithms: the feature selection wrapper Sequential Forward Elimination (SFE), and two feature filter algotithms: Principal Component Analysis (PCA) and Kernel Principal Component Analysis (KPCA). Current Results of demonstrate that the Naïve Bayes classifier combined with KPCA produces the best overall results. Also confirming the efficiency os features. / Avaliação de classificadores na classificação de radiografias de tórax para o diagnóstico de pneumonia infantil Este trabalho dá continuidade ao Sistema de Auxílio a Diagnóstico chamado de PneumoCAD para a detecção de pneumonia infantil por meio de imagens radiográficas, com o objetivo de aprimorar a acurácia, robustez e testar as características extraídas anteriormente. Nós implementamos cinco classificadores contemporâneos, sendo estes: Naïve Bayes, K-Nearest Neighbor (KNN), Support Vector Machines (SVM), Multi-Layer Perceptron (MLP) e Árvore de decisão. Combinamos os classificadores com três algoritmos de redução de dimensionalidade: o wrapper Sequential Forward Elimination (SFE) e dois filtros: Principal Component Analysis (PCA) e Kernel Principal Component Analysis (KPCA). Os resultados atuais mostram que o Naïve Bayes combinado com o KPCA produzem o melhor resultado (96% de acurácia). Também confirmando a eficiência das características.

Este trabalho teve como objetivo desenvolver uma base de imagens mamográficas online com acesso público para desenvolvimento, testes e avaliação comparativa de esquemas computadorizados de auxilio ao diagnóstico (CADs). A base contem imagens de vários hospitais, com grande variedade de laudos, também disponíveis na base assim como informações sobre dados clínicos (não confidenciais) dos pacientes. Uma interface detalhada foi criada para permitir o fácil acesso público, permitindo o uso de ferramentas de busca, recorte, analise estatística e inserção remota de imagens, entre outras. Testes comparativos com bases já existentes e amplamente usadas mostraram que a base desenvolvida tem quantidade e qualidade de imagens comparável ou superior as outras, além de oferecer uma quantidade de ferramentas muito maior. / This work has the objective of developing an online mamographic image database with public access for development, test and evaluation of computer-aided diagnosis (CAD). The database contains images from several hospitals, with great variety of medical reports, also available on the database together with other clinical information (not classified) from the patient. A detailed interface was developed to allow easy public access, allowing the use of tools for search, clipping, statistical analyses, remote insertion and others. Comparative test with other already existing databases shown that the presented database has quantity and quality comparable or superior to the others and offers a greater set of tools for the user.

Neste trabalho, a mineração de regras de associação é utilizada para dar suporte a dois tipos de sistemas médicos: os sistemas de busca por conteúdo em imagens (Content-based Image Retrieval - CBIR) e os sistemas de auxílio ao diagnóstico (Computer Aided Diagnosis - CAD). Na busca por conteúdo, regras de associação são empregadas para reduzir a dimensionalidade dos vetores de características que representam as imagens e para diminuir o ``gap semântico\'\', que existe entre as características de baixo nível das imagens e seu significado semântico. O algoritmo StARMiner (Statistical Association Rule Miner) foi desenvolvido para associar características de baixo nível das imagens com o seu significado semântico, sendo também utilizado para realizar seleção de características em bases de imagens médicas, melhorando a precisão dos sistemas CBIR. Para dar suporte aos sistemas CAD, o método IDEA (Image Diagnosis Enhancement through Association rules) foi desenvolvido. Nesse método regras de associação são empregadas para sugerir uma segunda opinião ou diagnóstico preliminar de uma nova imagem para o radiologista. A segunda opinião automaticamente gerada pelo método pode acelerar o processo de diagnóstico de uma imagem ou reforçar uma hipótese, trazendo ao especialista médico um apoio estatístico da situação sendo analisada. Dois novos algoritmos foram propostos: um para pré-processar as características de baixo nível das imagens médicas e, o outro, para propor diagnósticos baseados em regras de associação. Vários experimentos foram realizados para validar os métodos desenvolvidos. Os experimentos realizados indicam que o uso de regras de associação pode contribuir para melhorar a busca por conteúdo e o diagnóstico de imagens médicas, consistindo numa poderosa ferramenta para descoberta de padrões em sistemas médicos / In this work we take advantage of association rule mining to support two types of medical systems: the Content-based Image Retrieval (CBIR) and the Computer-Aided Diagnosis (CAD) systems. For content-based retrieval, association rules are employed to reduce the dimensionality of the feature vectors that represent the images and to diminish the semantic gap that exists between low-level features and its high-level semantical meaning. The StARMiner (Statistical Association Rule Miner) algorithm was developed to associate low-level features with their semantical meaning. StARMiner is also employed to perform feature selection in medical image datasets, improving the precision of CBIR systems. To improve CAD systems, we developed the IDEA (Image Diagnosis Enhancement through Association rules) method. Association rules are employed to suggest a second opinion to the radiologist or a preliminary diagnosis of a new image. A second opinion automatically obtained can accelerate the process of diagnosing or strengthen a hypothesis, giving to the physician a statistical support to the decision making process. Two new algorithms are developed to support the IDEA method: to pre-process low-level features and to propose a diagnosis based on association rules. We performed several experiments to validate the developed methods. The results indicate that association rules can be successfully applied to improve CBIR and CAD systems, empowering the arsenal of techniques to support medical image analysis in medical systems

Lung cancer is the leading cause of cancer death in the United States. While low-dose computed tomography (CT) screening reduces lung cancer mortality by 20%, 97% of suspicious lesions are found to be benign upon further investigation. Computer-aided diagnosis (CAD) tools can improve the accuracy of CT screening, however, current CAD tools which focus on imaging characteristics of the nodule alone are challenged by the limited data captured in small, early identified nodules. We hypothesize a CAD tool that incorporates quantitative CT features from the surrounding lung parenchyma will improve the ability of a CAD tool to determine the malignancy of a pulmonary nodule over a CAD tool that relies solely on nodule features. Using a higher resolution research cohort and a retrospective clinical cohort, two CAD tools were developed with different intentions. The research-driven CAD tool incorporated nodule, surrounding parenchyma, and global lung measurements. Performance was improved with the inclusion of parenchyma and global features to 95.6%, compared to 90.2% when only nodule features were used. The clinically-oriented CAD tool incorporated nodule and parenchyma features and clinical risk factors and identified several features robust to CT variability, resulting in an accuracy of 71%. This study supports our hypothesis that the inclusion of parenchymal features in the developed CAD tools resulted in improved performance compared to the CAD tool constructed solely with nodule features. Additionally, we identified the optimal amount of lung parenchyma for feature extraction and explored the potential of the CAD tools in a clinical setting.

Over the years, exemplar-based methods have yielded significant improvements over their model-based counterparts in image synthesis applications. Notably, texture synthesis algorithms using an exemplar-based approach have shown success where traditional stochastic methods failed. As an illustrative example, I will present an exemplar-based approach that yields substantial benefits for user-guided terrain synthesis using Digital Elevation Models (DEMs). This success is realized through exploitation of structural properties of natural terrain. In addition to their proliferation in the image synthesis domain, as annotated image datasets become increasingly available, exemplar-based methods are also gaining in popularity for image analysis applications. This thesis addresses the intersection between exemplar-based analysis and the problem of content-based image retrieval (CBIR). A basic problem in CBIR is the process by which the search criteria are refined by the user through the manipulation of returned exemplars. Exemplar-based analysis is particularly well-suited to query refinement due to its interpretability and the ease with which it can be incorporated into an interactive system. I investigate this connection in the domain of Computer-Assisted Diagnosis (CAD) of dermatological images. I demonstrate that exemplar-based approaches in CBIR can be effective for diagnosing pigmented skin lesions (PSLs). I will present an exemplar-based algorithm for segmenting PSLs in dermatoscopic images. In addition, I will present a generalized representation of dermoscopic features for detection and matching. This representation not only leads to an exemplar-based PSL diagnosis scheme, but it also enables us to realize interactive region-of-interest retrieval, which includes a relevance feedback mechanism to facilitate more flexible query-by-example analysis. Finally, I will assess the benefit of this CBIR-CAD approach through both quantitative evaluations and user studies.

Introduction: A computer-based classification system is proposed for the characterization of cervical intervertebral disc degeneration from saggital magnetic resonance images. Materials and methods: Cervical intervertebral discs from saggital magnetic resonance images where assessed by an experienced orthopaedist as normal or degenerated (narrowed) employing Matsumoto’s classification scheme. The digital images where enhanced and the intervertebral discs which comprised the regions of interest were segmented. First and second order statistics textural features extracted from thirty-four discs (16 normal and 16 degenerated) were used in order to design and test the classification system. In addition textural features were calculated employing Laws TEM images. The existence of statistically significant differences between the textural features values that were generated from normal and degenerated discs was verified employing the Student’s paired t-test. A subset with the most discriminating features (p<0.01) was selected and the Exhaustive Search and Leave-One-Out methods were used to find the best features combination and validate the classification accuracy of the system. The proposed system used the Least Squares Minimum Distance Classifier in combination with four textural features with comprised the best features combination in order to classify the discs as normal or degenerated. Results: The overall classification accuracy was 93.8% misdiagnosing 2 discs. In addition the system’s sensitivity in detecting a narrow disc was 93.8% and its specificity was also 93.8%. Conclusion: Further investigation and the use of a larger sample for validation could make the proposed system a trustworthy and useful tool to the physicians for the evaluation of degenerative disc disease in the cervical spine. / Σκοπός: Η στένωση των μεσοσπονδύλιων δίσκων της αυχενικής μοίρας, ως κύρια έκφραση εκφυλιστικής νόσου, είναι μια από τις σημαντικότερες αιτίες πρόκλησης πόνου στην περιοχή του αυχένα. Στην κλινική πράξη η αξιολόγηση της στένωσης γίνεται μέσω μέτρησης του μεσοσπονδύλιου διαστήματος, σε διάφορες απεικονίσεις της αυχενικής μοίρας του ασθενούς. Στην παρούσα εργασία προτείνεται μια υπολογιστική μέθοδος ανάλυσης εικόνας, για την αυτοματοποιημένη εκτίμηση της στένωσης από εικόνες μαγνητικής τομογραφίας. Υλικό και Μέθοδος: Μελετήθηκαν 34 μεσοσπονδύλιοι δίσκοι από οβελιαίες τομές μαγνητικής τομογραφίας της αυχενικής μοίρας, οι οποίες ελήφθησαν με χρήση Τ2 ακολουθίας. Η στένωση των μεσοσπονδύλιων δίσκων αξιολογήθηκε από έμπειρο ορθοπαιδικό βάσει της κλίμακας Matsumoto. Οι δίσκοι χωρίστηκαν σε δύο κατηγορίες: (α) 16 φυσιολογικοί και (β) 16 δίσκοι που παρουσίαζαν στένωση. Με χρήση διαδραστικού περιβάλλοντος επεξεργασίας εικάνας καθορίστηκε το περίγραμμα των μεσοσπονδύλιων δίσκων οι οποίοι αποτελούν τις προς ανάλυση περιοχές ενδιαφέροντος (Π.Ε.). Σε κάθε Π.Ε. εφαρμόστηκαν αλγόριθμοι εξαγωγής χαρακτηριστικών υφής. Συγκεκριμένα υπολογίστικαν χαρακτηριστικά υφής από στατιστικά πρώτης και δεύτερης τάξης καθώς και χαρακτηριστικά από τα μέτρα ενέργειας υφλης κατλα Laws. Τα παραπάνω χαρακτηριστικά, ποσοτικοποιούν διαγνωστικές πληροφορίες της έντασης του σήματος της Π.Ε. και συσχετίζονται με τη βιοχημική σύσταση των απεικονιζόμενων δομών. Τα εξαχθέντα χαρακτηριστικά υφής αξιοποιήθηκαν για τη σχεδίαση του ταξινομητή ελάχιστης απόστασης ελαχίστων τετραγώνων, ο οποίος χρησιμοποιήθηκε για το διαχωρισμό μεταξύ φυσιολογικών δίσκων και δίσκων που παρουσίαζαν στένωση (εκφυλισμένων). Αποτελέσματα: Η ακρίβεια της ταξινόμησης φυσιολογικών και εκφυλισμένων μεσοσπονδύλιων δίσκων ανήλθε σε 93.8%. Η ευαισθησία καθώς και η ειδικότητα της μεθόδου, σε ότι αφορά την ανίχνευση εκφυλισμένων δίσκων, είναι επίσης 93.8%. Συμπέρασμα: Με δεδομένο το μικρό μέγεθος του δείγματος που χρησιμοποιήθηκε για το σχεδιασμό της μεθόδου, απαιτούνται περετέρω εργασίες πιστοποίησης της ακρίβειας ταξινόμησης, προκειμένου η μέθοδος αυτή να αξιοποιηθεί από ακτινολόγους και ορθοπαιδικους, ως βοηθητικό διαγνωστικό εργαλείο.

Este trabalho apresenta o estudo e implementação de um banco de conhecimentos para auxiliar o diagnóstico de lesões da mama por inspeção visual, permitindo ao médico consultas através de características pictóricas da imagem e a comparação visual entre imagem investigada e imagens previamente classificadas e suas informações clínicas. As imagens encontram-se classificadas no banco de conhecimentos segundo o padrão \"Breast imaging reporting and data systems\" (BI-RADS) do Colégio Americano de Radiologia. A seleção das imagens, informações clínicas representativas, bem como sua classificação foram realizada em conjunto com médicos radiologistas do Centro de Ciências das Imagens e Física Médica (CCIFM) da Faculdade de Medicina de Ribeirão Preto (FMRP) da Universidade de São Paulo (USP). O processo de indexação e recuperação das imagens é baseado em atributos de textura extraídos de \"Regions of interest\" (ROIs) previamente estabelecidas em mamogramas digitalizados. Para simplificar este processo, foi utilizado a Análise de Componentes Principais (PCA), que visa a redução do número de atributos de textura e as informações redundantes existentes. Os melhores resultados obtidos foram para as ROIs 139 (Precisão = 0.80), 59 (Precisão = 0.86) e um valor de 100% de acerto para a ROI 40. / This work presents the survey and implementation of a database of knowledge to aid the diagnostic of breast lesions through visual inspection, allowing the physician a seach through the characteristics of the contents of the image and the visual comparison between the analysed image and the previously classified images and its clinical information. The images are classified into the database of knowledge according to the pattern Breast Imaging Reporting and Data Systems (BI-RADS) of the American College of Radiology. The selection of the images, the representative clinical information, as well as its classification have been performed in conjunction with practictioners radiologists of the Centro de Ciências das Imagens e Física Médica (CCIFM) from Faculdade de Medicina de Ribeirão Preto (FMRP) from Universidade de São Paulo (USP). The process of indexing and retrieving the images is based on characteristic of the texture extracted from the regions of interest (ROIs) previously established through scanned mammograms. To simplify this path, the Principal Components Analysis (PCA) was used it aims the reduction of the number of features of texture and the existing redundant information. The best results obtained were to the ROIs 139 (precision = 0.80), 59 (precision = 0.86) and a value of 100% of precision for ROI 40.

Les outils méthodologiques en indexation et classification des images par le contenu sont déjà assez matures et ce domaine s’ouvre vers les applications médicales. Dans cette thèse,nous nous intéressons à l'indexation visuelle, à la recherche et à la classification des images cérébrales IRM par le contenu pour l'aide au diagnostic de la maladie d'Alzheimer (MA). L'idée principale est de donner au clinicien des informations sur les images ayant des caractéristiques visuelles similaires. Trois catégories de sujets sont à distinguer: sujets sains (NC), sujets à troubles cognitifs légers (MCI) et sujets atteints par la maladie d'Alzheimer(AD). Nous représentons l’atrophie cérébrale comme une variation de signal dans des images IRM (IRM structurelle et IRM de Tenseur de Diffusion). Cette tâche n'est pas triviale,alors nous nous sommes concentrés uniquement sur l’extraction des caractéristiques à partir des régions impliquées dans la maladie d'Alzheimer et qui causent des changements particuliers dans la structure de cerveau : l'hippocampe le Cortex Cingulaire Postérieur. Les primitifs extrais sont quantifiés en utilisant l'approche sac de mots visuels. Cela permet de représenter l’atrophie cérébrale sous forme d’une signature visuelle spécifique à la MA.Plusieurs stratégies de fusion d’information sont appliquées pour renforcer les performances de système d’aide au diagnostic. La méthode proposée est automatique (sans l’intervention de clinicien), ne nécessite pas une étape de segmentation grâce à l'utilisation d'un Atlas normalisé. Les résultats obtenus apportent une amélioration par rapport aux méthodes de l’état de l’art en termes de précision de classification et de temps de traitement. / Content-Based Visual Information Retrieval and Classification on Magnetic Resonance Imaging (MRI) is penetrating the universe of IT tools supporting clinical decision making. A clinician can take profit from retrieving subject’s scans with similar patterns. In this thesis, we use the visual indexing framework and pattern recognition analysis based on structural MRIand Tensor Diffusion Imaging (DTI) data to discriminate three categories of subjects: Normal Controls (NC), Mild Cognitive Impairment (MCI) and Alzheimer's Disease (AD). The approach extracts visual features from the most involved areas in the disease: Hippocampusand Posterior Cingulate Cortex. Hence, we represent signal variations (atrophy) inside the Region of Interest anatomy by a set of local features and we build a disease-related signature using an atlas based parcellation of the brain scan. The extracted features are quantized using the Bag-of-Visual-Words approach to build one signature by brain/ROI(subject). This yields a transformation of a full MRI brain into a compact disease-related signature. Several schemes of information fusion are applied to enhance the diagnosis performance. The proposed approach is less time-consuming compared to the state of thearts methods, computer-based and does not require the intervention of an expert during the classification/retrieval phase.

Neste trabalho, a mineração de regras de associação é utilizada para dar suporte a dois tipos de sistemas médicos: os sistemas de busca por conteúdo em imagens (Content-based Image Retrieval - CBIR) e os sistemas de auxílio ao diagnóstico (Computer Aided Diagnosis - CAD). Na busca por conteúdo, regras de associação são empregadas para reduzir a dimensionalidade dos vetores de características que representam as imagens e para diminuir o ``gap semântico\'\', que existe entre as características de baixo nível das imagens e seu significado semântico. O algoritmo StARMiner (Statistical Association Rule Miner) foi desenvolvido para associar características de baixo nível das imagens com o seu significado semântico, sendo também utilizado para realizar seleção de características em bases de imagens médicas, melhorando a precisão dos sistemas CBIR. Para dar suporte aos sistemas CAD, o método IDEA (Image Diagnosis Enhancement through Association rules) foi desenvolvido. Nesse método regras de associação são empregadas para sugerir uma segunda opinião ou diagnóstico preliminar de uma nova imagem para o radiologista. A segunda opinião automaticamente gerada pelo método pode acelerar o processo de diagnóstico de uma imagem ou reforçar uma hipótese, trazendo ao especialista médico um apoio estatístico da situação sendo analisada. Dois novos algoritmos foram propostos: um para pré-processar as características de baixo nível das imagens médicas e, o outro, para propor diagnósticos baseados em regras de associação. Vários experimentos foram realizados para validar os métodos desenvolvidos. Os experimentos realizados indicam que o uso de regras de associação pode contribuir para melhorar a busca por conteúdo e o diagnóstico de imagens médicas, consistindo numa poderosa ferramenta para descoberta de padrões em sistemas médicos / In this work we take advantage of association rule mining to support two types of medical systems: the Content-based Image Retrieval (CBIR) and the Computer-Aided Diagnosis (CAD) systems. For content-based retrieval, association rules are employed to reduce the dimensionality of the feature vectors that represent the images and to diminish the semantic gap that exists between low-level features and its high-level semantical meaning. The StARMiner (Statistical Association Rule Miner) algorithm was developed to associate low-level features with their semantical meaning. StARMiner is also employed to perform feature selection in medical image datasets, improving the precision of CBIR systems. To improve CAD systems, we developed the IDEA (Image Diagnosis Enhancement through Association rules) method. Association rules are employed to suggest a second opinion to the radiologist or a preliminary diagnosis of a new image. A second opinion automatically obtained can accelerate the process of diagnosing or strengthen a hypothesis, giving to the physician a statistical support to the decision making process. Two new algorithms are developed to support the IDEA method: to pre-process low-level features and to propose a diagnosis based on association rules. We performed several experiments to validate the developed methods. The results indicate that association rules can be successfully applied to improve CBIR and CAD systems, empowering the arsenal of techniques to support medical image analysis in medical systems

O câncer de mama é hoje uma das principais causas de mortalidade de mulheres em todo o mundo. Porém, a sua detecção no estágio inicial de desenvolvimento aumenta consideravelmente as chances de cura. Exatamente por isso estão sendo desenvolvidos vários tipos de sistemas computacionais baseados em processamento de imagens em centros de pesquisas no mundo todo, a fim de auxiliar o radiologista na precisão do seu diagnóstico. A pesquisa aqui apresentada se insere nesse contexto e consistiu no desenvolvimento de um sistema computacional para detectar uma das estruturas que podem ser indício da presença do câncer de mama: os agrupamentos (\"clusters\") de microcalcificações. O sistema aqui apresentado tem como fonte de dados mamogramas digitalizados, nos quais são aplicadas técnicas de processamento para extrair as regiões de interesse e detectar os possíveis \"clusters\" existentes. Os resultados dos testes realizados mostraram que o sistema desenvolvido apresentou uma eficiência de 94% na identificação correta de \"clusters\". / Breast cancer is one of the main causes of women death all over the world. However, early detection of the disease increases greatly the possibility of cure. Therefore, several types of computer systems based on image processing are being developed by many research groups in order to aid the radiologist in the accuracy of the diagnosis. The work presented here is inserted in this context corresponding to the development of a computer system designed to detect microcalcifications clusters - structures which can be a strong indicative of breast cancer. This system database is digitized mammograms, to which image processing techniques are applied in order to detect regions of interest and the possible clusters. The results from the tests have shown an efficacy of 94% of the system in clusters correct identification.