Rozprawy doktorskie na temat „Computer-Aided Tomography”

This thesis, describes work whose principal goal has been to research and iden- tify an optimum algorithm for computing the Radon transform - optimum in the sense of minimum CPU time for maximum image fidelity. This transform is the basis for a wide range of applications which are either directly or indirectly associ- ated with Computer Aided Tomography. It has therefore been important to try and identify an optimum computational technique for the evaluation of this transform. A central theme to the computational techniques reported in this thesis has been the digital image rotation for which a new algorithm has been designed based en- tirely on integer arithmetic. This algorithm has been used for the evaluation of the forward and inverse Radon transforms (in particular, forward and back-projection). The relationship between the interpolation methods used for image rotation and the filtering techniques required to compute the inverse Radon transform has been stud- ied in detail. The result of this study has been to identify a specific data processing path which provides an algorithm whose 'computational energy' is approximately half that of previously published algorithms. With an optimum algorithm identified, research has been undertaken into the use of the Radon transform for two-dimensional array processing. It is shown that the Radon transform can be used to reduce the dimensionality of a problem from two-dimensions to a one-dimensional problem. By applying signal processors to the set of projections obtained by taking the Radon transform of an image, the inverse Radon transform can be used to reconstruct the processed image. The value of this approach lies in the fact that some useful one-dimensional signal processing algorithms do not have a two-dimensional extension or else the implementation of the two-dimensional algorithm is computationally intensive. An example of the latter case is reported.

Coronary artery disease is one of the most pernicious diseases around the world and early identification of vascular disease can help to reduce morbidity and mortality. Assessment of the degree of vascular obstruction, or stenosis, is critical for classifying the risks of the future vascular events. Automatic detection and quantification of stenosis are important in assessing coronary artery disease from medical imagery, especially for disease progression. Important factors affecting the reproducability and robustness of accuarate quantification arise from the partial volume effect and other noise sources. The main goal of this study is to present a fully automatic approach for detection and quantification of the stenosis in the coronary arteries. The proposed approach begins by building a 3D reconstruction of the coronary arterial system and then making accurate measurement of the vessel diameter from a robust estimate of the vessel cross-section. The proposed algorithm models the partial volume effect using a Markovian fuzzy clustering method in the process of accurate quantification of the degree of stenosis. To evaluate the accuracy and reproducibility of the measurement, the method was applied to a vascular phantom that was scanned using different protocols. The algorithm was applied to 20 CTA patient datasets containing a total of 85 stenoses, which were all successfully detected, with an average false positive rate of 0.7 per scan.

Good soil structure is important for sustaining agricultural production and preserving functions of the soil ecosystem. Soil aggregation is a critically important component of soil structure. Stable aggregates enable water infiltration, gas exchange for biological activities of plant roots and microorganisms, living space and surfaces for soil microbes, and contribute to stabilization of organic matter and storage of organic carbon (OC) in soil. Soil aggregation process involves fine roots, organic matter and fungal hyphae. Hyphal proliferation is essential for soil aggregation, and sequestration of OC in soil. Organic materials and available phosphorus are two of the major factors that influence fungi in soil. Organic materials are a source of energy for saprotrophic microbes and fungal hyphae may increase in the presence of organic matter. Phosphorus is an important element usually found depleted in soil ecosystems. The low availability of phosphorus may limit the biological activity of microbes. Arbuscular mycorrhizal fungi (AM) benefit plants by delivering phosphorus to the root system. However, the density and the length of hyphae of AM fungi do not appear to be increased by available phosphorus. We do not yet have a mechanism to directly quantify the density of hyphae in soil. A number of indirect methods have been used to visualize distribution of fungi in soil. Reliable analyses of soil are limited because of the use of destructive harvests to gain information. Soils are fragile, and fragility limits opportunity for non-destructive analysis. The soil ecosystem is also complex. Soil particles are dense and the density obscures the visualization of fungal hyphae. Fungal hyphae are relatively fine and information at the small scale (<250 µm) is key to understanding how fungi respond to environmental stimuli. The experiments conducted here developed a new method for the observation of fungi and quantification of hyphae in three dimensions. The methods were first tested using a melanised saprotrophic fungus, 222A. The response of two AM fungi, Glomus mosseae and Glomus pellucidum, growing together to common environmental factors was then quantified. Hyphae were quantified in an artificial soil matrix over 6 week’s incubation using micro-computer aided tomography (microCT). MicroCT provides three dimensional images of mycelia within electron lucent materials and enables the visualization and quantification of hyphae. Starch stimulated proliferation of 222A after 2 weeks. Starch, and starch and K2HPO4, stimulated proliferation of hyphae of AM fungi after 3 and 6 weeks. K2HPO4 stimulated hyphal proliferation only after 3 weeks. The images also indicate that fungal hyphae grew appressed to the surfaces of the particles rather than through the spaces between the particles. The capacity to quantify hyphae in three-dimensional space allows a wide range of questions to now be addressed. Apart from studying mechanisms of carbon turnover, more complex processes may now be considered. Soil is commonly thought of as a black box. That black box potentially is now a shade of grey.

The present work expose a technique of estimation of optimal direction for placing dental implant. A volumetric computed tomography (CT) scan is used as a help of the following searches. The work offers criteria of the optimal implant placement direction and methods of evaluation on direction’s significance. The technique utilizes structure tensor to find a normal to the jawbone surface. Direction of that normal is then used as initial direction for search of optimal direction.

The technique described in the present work aimed to support doctor’s decisions during dental implantation treatment.

Medical imaging informatics is one of the important research areas in radiology that studies how information available on medical images is retrieved, analyzed, and enhanced. Recent development in medical imaging informatics has resulted in improvement of diagnostic accuracy based on imaging examinations, as well as efficiency in clinical workflow. Computer aided diagnosis (CAD) and electronic patient record system (ePR) are both topics in medical imaging informatics that have matured from research concepts into commercially available computerized systems in clinical environment. The current challenges are to further broaden their scope of applications. In this thesis project, I developed a CAD system for interpreting PET/CT examinations and an ePR system for patient data integration in neurosurgery suites. Specifically, the CAD system in this project was designed to automatically diagnose nasopharyngeal carcinoma (NPC) on Positron emission tomography/computed tomography (PET/CT) examinations, which aimed to detect and classify both the primary NPC and its nodal metastasis. The regions of interests (ROIs) were segmented from the PET images and registered onto the CT in order to combine the imaging features from both modalities and the a priori anatomical knowledge of the suspicious lesion. These combined features were then classified by a support vector machine (SVM) to generate the final diagnosis result. The system was validated with 25 PET/CT examinations from 10 patients suffering from NPC, and the result produced by the system was compared to the gold standard of lesions manually contoured by experienced radiologists. The results confirmed that the system successfully distinguished all 53 genuine lesions from the mimickers due to normal physiological uptake and artifacts that also produced potentially confusing signals. The second part of the project involved development of an electronic patient record system (ePR) that integrated all the myriad of images and different types of clinical information before, during, and after neurosurgery operations, in order to enhance efficiency of work flow in this unique clinical environment. The system comprises of pre-, intra-, and post-operation modules which correspond to the different stages of the neurosurgery. The pre-op module was developed to store and categorize all images and data before the procedure to assist the surgeons in planning operation. The intra-op module integrates all the input signals, waveforms, images and videos that are produced by different imaging and physiological monitoring devices in the operation room during the surgery, and displays all the relevant information in a single large screen in real time to ease monitoring of the procedure. The post-op module helps surgeons to review all the data acquired from all the prior stages for follow-up of the treatment outcome. One-tumor case was utilized to test the pre-op module, and the signals and waveforms simulators were used to evaluate the performance of the intra-op module. In summary, two different medical informatics systems, a CAD and an ePR system were developed. Both showed promising results in laboratory tests. Future work would involve performance enhancement and feedback of the systems, and ultimately evaluation of these systems in the clinical environment.
published_or_final_version
Diagnostic Radiology
Master
Master of Philosophy

Environ 150.000 personnes souffrent en France d'une épilepsie partielle réfractaire à tous les médicaments. La chirurgie, qui constitue aujourd’hui le meilleur recours thérapeutique nécessite un bilan préopératoire complexe. L'analyse de données d'imagerie telles que l’imagerie par résonance magnétique (IRM) anatomique et la tomographie d’émission de positons (TEP) au FDG (fluorodéoxyglucose) tend à prendre une place croissante dans ce protocole, et pourrait à terme limiter de recourir à l’électroencéphalographie intracérébrale (SEEG), procédure très invasive mais qui constitue encore la technique de référence. Pour assister les cliniciens dans leur tâche diagnostique, nous avons développé un système d'aide au diagnostic (CAD) reposant sur l'analyse multivariée de données d'imagerie. Compte tenu de la difficulté relative à la constitution de bases de données annotées et équilibrées entre classes, notre première contribution a été de placer l'étude dans le cadre méthodologique de la détection du changement. L'algorithme du séparateur à vaste marge adapté à ce cadre là (OC-SVM) a été utilisé pour apprendre, à partir de cartes multi-paramétriques extraites d'IRM T1 de sujets normaux, un modèle prédictif caractérisant la normalité à l'échelle du voxel. Le modèle permet ensuite de faire ressortir, dans les images de patients, les zones cérébrales suspectes s'écartant de cette normalité. Les performances du système ont été évaluées sur des lésions simulées ainsi que sur une base de données de patients. Trois extensions ont ensuite été proposées. D'abord un nouveau schéma de détection plus robuste à la présence de bruit d'étiquetage dans la base de données d'apprentissage. Ensuite, une stratégie de fusion optimale permettant la combinaison de plusieurs classifieurs OC-SVM associés chacun à une séquence IRM. Enfin, une généralisation de l'algorithme de détection d'anomalies permettant la conversion de la sortie du CAD en probabilité, offrant ainsi une meilleure interprétation de la sortie du système et son intégration dans le bilan pré-opératoire global
One third of patients suffering from epilepsy are resistant to medication. For these patients, surgical removal of the epileptogenic zone offers the possibility of a cure. Surgery success relies heavily on the accurate localization of the epileptogenic zone. The analysis of neuroimaging data such as magnetic resonance imaging (MRI) and positron emission tomography (PET) is increasingly used in the pre-surgical work-up of patients and may offer an alternative to the invasive reference of Stereo-electro-encephalo -graphy (SEEG) monitoring. To assist clinicians in screening these lesions, we developed a computer aided diagnosis system (CAD) based on a multivariate data analysis approach. Our first contribution was to formulate the problem of epileptogenic lesion detection as an outlier detection problem. The main motivation for this formulation was to avoid the dependence on labelled data and the class imbalance inherent to this detection task. The proposed system builds upon the one class support vector machines (OC-SVM) classifier. OC-SVM was trained using features extracted from MRI scans of healthy control subjects, allowing a voxelwise assessment of the deviation of a test subject pattern from the learned patterns. System performance was evaluated using realistic simulations of challenging detection tasks as well as clinical data of patients with intractable epilepsy. The outlier detection framework was further extended to take into account the specificities of neuroimaging data and the detection task at hand. We first proposed a reformulation of the support vector data description (SVDD) method to deal with the presence of uncertain observations in the training data. Second, to handle the multi-parametric nature of neuroimaging data, we proposed an optimal fusion approach for combining multiple base one-class classifiers. Finally, to help with score interpretation, threshold selection and score combination, we proposed to transform the score outputs of the outlier detection algorithm into well calibrated probabilities

L'évolution des techniques d'acquisition des imageries médicales et leur importance de plus en plus grande dans la prise en charge du patient (diagnostic, préparation d'intervention, suivi, etc.), font émerger de nouveaux besoins autour du traitement informatique des images. La reconnaissance du type de lésions hépatiques est un grand enjeu, notamment car le cancer du foie, létal et très répandu, est souvent diagnostiqué trop tard pour sauver le patient. C'est dans ce cadre qu'est né le projet de recherche de ce manuscrit, fruit d'une collaboration entre la société IMAIOS et le Laboratoire d'Informatique, de Robotique et de Microélectronique de Montpellier (LIRMM).Cette thèse présente un système complet et automatique permettant, à partir d'images de lésions au format médical DICOM, d'extraire des descripteurs visuels de divers nodules hépatiques puis de les différencier à l'aide de ces derniers. Les contributions décrites s'articulent autour de divers axes : normalisation des niveaux de gris des images de lésions par rapport au foie sain, proposition, analyse et tests de descripteurs visuels (s'appuyant notamment sur les informations temporelles ou de densité des tissus), caractérisations diverses des différents types de lésions grâce à ces descripteurs et à un algorithme de classification. Les données sur lesquelles ces travaux ont été effectués sont des examens scanner multiphasiques
Medical imaging acquisition has taken benefits from recent advances and is becoming more and more important in the patient care process. New needs raise, which are related to image processing. Hepatic lesion recognition is a hot topic, especially because liver cancer is wide-spread and leads to death, most of the time because of the diagnosis which is made too late. In this context is born this manuscrit research project, a collaboration between IMAIOS company and the Laboratory of Informatics, Robotics and Micro-electronics ofMontpellier (LIRMM).This thesis presents a complete and automated system that extracts visual features from lesion images in the medical format DICOM, then differenciate them on these features.The various described contributions are: intensity normalization using healthy liver values, analysis and experimentations around new visual features, which use temporal information or tissue density, different kind of caracterisation of the lesions. This work has been done on multi-phase Computed Tomography acquisitions

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.

Aim of this work is to develop and to validate a structured methodology to investigate the three-dimensional variation that occurs around implants in dentistry. Surgeons need to know in an objective way if what they are doing is correct and if it is the best for the patient. In last decades implantology deeply changed the way to operate of dentists, giving to the patients new opportunities to replace missing teeth. Implantology has known a very big spread all around the world and numbers of patients treated with success is growing year by year. To know exactly what happens around implants is a growing need for clinicians. A standardized method that can investigate in an objective way what soft and hard tissues do around implants doesn’t exist yet. The solutions that researchers used in literature are various and difficult to compare each other. This work after a general discussion that follows the evolution of implantology, wants to investigate some new instruments that could lend to the comparability of results among different studies and finally to give better answers to the clinical questions. Using the method proposed in this work, soft-hard tissue variation are been evaluated from a new prospective that gave impressive results both qualitatively and quantitatively speaking. The procedure is recommended as a new aid in the future studies.
Obiettivo del lavoro è di sviluppare e validare una metodologia strutturata per indagare la variazione tridimensionale che avviene intorno agli impianti endossei in odontoiatria. I chirurghi hanno bisogno di sapere in modo oggettivo se quello che stanno facendo è corretto ed è la migliore terapia per il paziente. Negli ultimi decenni l’implantologia ha profondamente cambiato il modo di operare dei dentisti, dando ai pazienti nuove opportunità per sostituire i denti mancanti. Implantologia ha conosciuto una grande diffusione in tutto il mondo e il numero di pazienti trattati con successo sta crescendo di anno in anno. Sapere esattamente ciò che accade intorno agli impianti è una crescente necessità per i medici. Un metodo standardizzato che possa indagare in modo oggettivo come si modifichino i tessuti duri e molli intorno agli impianti non esiste ancora. Le soluzioni che i ricercatori hanno utilizzato in letteratura sono molteplici e difficili da confrontare tra loro. Questo lavoro, dopo una discussione generale che segue l'evoluzione dell’implantologia, vuole approfondire l’uso di alcuni nuovi strumenti che possano portare alla comparabilità dei risultati tra i diversi studi e, infine, di dare risposte migliori alle domande cliniche che ancora non hanno risposta. Utilizzando il metodo proposto in questo lavoro, è possibile valutare i tessuti peri-implantari da una nuova prospettiva che ha dato risultati impressionanti sia sul versante qualitativo sia su quello quantitativo. La procedura è un ausilio raccomandato come nuovo aiuto nei futuri studi.

Dans ce projet de recherche financé en contrat CIFRE avec GE Healthcare, on cherche a prédire les épisodes cardio-vasculaire adverses majeurs (ECAM), c’est à dire typiquement les embolies et les anévrismes dans l’aorte et les artères coronaires, qui donnent lieu a une respectivement à une interruption catastrophique du flux sanguin vers le coeur et donc un infarctus, ou à une hémorragie interne. Les deux types d’épisodes sont extrêmement graves. Lorsqu’un patient est hospitalisé pour une alerte reliée à ces épisodes, il va subir un examen scanner X, injecté ou non, plus ou moins invasif. Un objectif majeur de cette recherche est d’utiliser au mieux l’information obtenue sous forme d’images 3D ainsi que l’historique du patient pour éviter de soumettre le patient à des examens inutiles, invasifs ou dangereux, tout en garantissant le meilleur résultat clinique. Les méthodologies proposées reposeront sur des techniques d’analyse et traitement d’image, de vision par ordinateur et d’imagerie médicale qui seront développée en partenariat entre GE Healthcare et le laboratoire Centre de Vision Numérique (CVN) de CentraleSupélec
This research project is expected to be financed by a CIFRE scholarship in collaboration between GE Healthcare and CentraleSupelec. We are seeking to predict Major Adverse Cardiovascular Events (MACE). These are typically embolism and aneurisms in the aorta and the coronary arteries, that give rise respectively to interrupted blood flow to the heart and so a risk of infarctus, or major hemorrhage. Both are life-threatening. When a patient is brought to hospital for an alert (angina, etc), they will undergo an X-ray CAT scan, which can be more or less invasive. A major objective of this research is to utilize as well as possible the available information in the form of 3D images together with patient history and other data, in order to avoid needless, invasive, irradiating or dangerous exams, while simultaneously guaranteeing optimal care and the best possible clinical outcome. The proposed methodologies include image analysis, image processing, computer vision and medical imaging procedures and methods, that will be developed in partnership between GE Healthcare and the CVN lab of CENTRALE SUPELEC

La tomographie par émission de positons (TEP) est une méthode d'imagerie clinique en forte expansion dans le domaine de l'oncologie. De nombreuses études cliniques montrent que la TEP permet, d'une part de diagnostiquer et caractériser les lésions cancéreuses à des stades plus précoces que l'imagerie anatomique conventionnelle, et d'autre part d'évaluer plus rapidement la réponse au traitement. Le raccourcissement du cycle comprenant le diagnostic, la thérapie, le suivi et la réorientation thérapeutiques contribue à augmenter le pronostic vital du patient et maîtriser les coûts de santé. La durée d'un examen TEP ne permet pas de réaliser une acquisition sous apnée. La qualité des images TEP est par conséquent affectée par les mouvements respiratoires du patient qui induisent un flou dans les images. Les effets du mouvement respiratoire sont particulièrement marqués au niveau du thorax et de l'abdomen. Plusieurs types de méthode ont été proposés pour corriger les données de ce phénomène, mais elles demeurent lourdes à mettre en place en routine clinique. Des travaux récemment publiés proposent une évaluation de ces méthodes basée sur des critères de qualité tels que le rapport signal sur bruit ou le biais. Aucune étude à ce jour n'a évalué l'impact de ces corrections sur la qualité du diagnostic clinique. Nous nous sommes focalisés sur la problématique de la détection des lésions du thorax et de l'abdomen de petit diamètre et faible contraste, qui sont les plus susceptibles de bénéficier de la correction du mouvement respiratoire en routine clinique. Nos travaux ont consisté dans un premier temps à construire une base d'images TEP qui modélisent un mouvement respiratoire non-uniforme, une variabilité inter-individuelle et contiennent un échantillonnage de lésions de taille et de contraste variable. Ce cahier des charges nous a orientés vers les méthodes de simulation Monte Carlo qui permettent de contrôler l'ensemble des paramètres influençant la formation et la qualité de l'image. Une base de 15 modèles de patient a été créée en adaptant le modèle anthropomorphique XCAT sur des images tomodensitométriques (TDM) de patients. Nous avons en parallèle développé une stratégie originale d'évaluation des performances de détection. Cette méthode comprend un système de détection des lésions automatisé basé sur l'utilisation de machines à vecteurs de support. Les performances sont mesurées par l'analyse des courbes free-receiver operating characteristics (FROC) que nous avons adaptée aux spécificités de l'imagerie TEP. L'évaluation des performances est réalisée sur deux techniques de correction du mouvement respiratoire, en les comparant avec les performances obtenues sur des images non corrigées ainsi que sur des images sans mouvement respiratoire. Les résultats obtenus sont prometteurs et montrent une réelle amélioration de la détection des lésions après correction, qui approche les performances obtenues sur les images statiques.

Dans le cadre de cette thèse, nous nous sommes intéressés à l’étude de l’apport d’une aide assistée par ordinateur au diagnostic de certaines maladies dégénératives du cerveau, en explorant les images de tomographie par émission de positons, par des techniques de traitement d’image et d’analyse statistique.Nous nous sommes intéressés à la représentation corticale des 116 régions anatomiques, en associant à chacune d’elles un vecteur d’attribut issu du calcul des 4 premiers moments des intensités de voxels, et en y incluant par ailleurs l’entropie. Sur la base de l’aire de courbes ROC, nous avons établi qualitativement la pertinence de chacune des régions anatomiques, en fonction du nombre de paramètres du vecteur d’attribut qui lui était associé, pour séparer le groupe des sujets sains de celui des sujets atteints de la maladie d’Alzheimer. Dans notre étude nous avons proposé une nouvelle approche de sélection de régions les plus pertinentes, nommée "combination matrix", en se basant sur un système combinatoire. Chaque région est caractérisée par les différentes combinaisons de son vecteur d’attribut. L’introduction des régions les plus pertinentes(en terme de pouvoir de séparation des sujets) dans le classificateur supervisé SVM nous a permis d’obtenir, malgré la réduction de dimension opérée, un taux de classification meilleur que celui obtenu en utilisant l’ensemble des régions
Our research focuses on presenting a novel computer-aided diagnosis technique for brain Positrons Emission Tomography (PET) images. It processes and analyzes quantitatively these images, in order to better characterize and extract meaningful information for medical diagnosis. Our contribution is to present a new method of classifying brain 18 FDG PET images. Brain images are first segmented into 116 Regions Of Interest (ROI) using an atlas. After computing some statistical features (mean, standarddeviation, skewness, kurtosis and entropy) on these regions’ histogram, we defined a Separation Power Factor (SPF) associated to each region. This factor quantifies the ability of each region to separate neurodegenerative diseases like Alzheimer disease from Healthy Control (HC) brain images. A novel region-based approach is developed to classify brain 18FDG-PET images. The motivation of this work is to identify the best regional features for separating HC from AD patients, in order to reduce the number of features required to achieve an acceptable classification result while reducing computational time required for the classification task

A análise de imagens médicas auxiliada por computador permite a análise quantitativa das anormalidades e garante maior precisão diagnóstica. Esse tipo de análise é importante para medicina nuclear com Single Photon Emission Computed Tomography (SPECT), pois no grupo de dados tridimensionais de imagens, padrões sutis de anormalidades muitas vezes são importantes achados clínicos. Porém, as imagens podem sofrer interferência de artefatos de atenuação da emissão de fótons por partes moles corporais, o que reduz sua acurácia diagnóstica. Desde que se possuam parâmetros de atenuação computados em um modelo que permita a comparação com imagens de um dado paciente, a interferência dos artefatos pode ser corrigida com ganho na acurácia diagnóstica, sem a necessidade de utilização de técnicas de correção que aumentem a dose de exposição à radiação pelo paciente. A proposta desse estudo foi a criação de um atlas de cintilografia de perfusão miocárdica, que foi obtido a partir de imagens de indíviduos normais, e o desenvolvimento de um algoritmo computacional para a detecção de anormalidades perfusionais miocárdicas, através da comparação estatística dos modelos do atlas com imagens de pacientes. Métodos de corregistro de imagens de mesma modalidade e outras técnicas de processamento de imagens foram estudados e utilizados para a comparação das imagens dos pacientes com o modelo apropriado. Pela análise visual dos modelos, verificou-se a sua validade como imagem representativa de normalidade perfusional. Para avaliação da detecção, a situação dos segmentos miocárdicos (normal ou anormal) indicada pelo algoritmo de detecção foi comparada com a situação apontada no laudo obtido pela concordância de dois especialistas, de modo a se verificar as concordâncias e discordâncias da técnica em relação ao laudo e se obter a significância estatística. Com isso, verificou-se um índice de concordância positiva da técnica em relação ao laudo de aproximadamente 50%, de concordância negativa próxima a 82% e de concordância geral próxima a 68%. O teste exato de Fisher foi aplicado às tabelas de contingência, obtendo-se um valor de p bicaudal inferior a 0,0001, indicando uma probabilidade muito baixa de as concordâncias terem sido obtidas pelo acaso. Melhorias no algoritmo deverão ser implementadas e testes futuros com um padrão-ouro efetivo serão realizados para validação da técnica.
The computer-aided medical imaging analysis allows the quantitative analysis of abnormalities and enhances diagnostic accuracy. This type of analysis is important for nuclear medicine that uses Single Photon Emission Computed Tomography (SPECT), because in the group of three-dimensional data images, subtle patterns of abnormalities often are important clinical findings. However, images can suffer interference from attenuation artifacts of the emission of photons by soft parts of the body, which reduces their diagnostic accuracy. Since there are attenuation parameters computed in a template that allows for comparison with images of a given patient, the artifacts interference can be corrected with a gain in diagnostic accuracy, without the need of using correction techniques that increase the radiation exposure dose of the patient. The purpose of this study was to create an atlas of myocardial perfusion scintigraphy, which was obtained from images of normal individuals and the development of a computational algorithm for detection of myocardial perfusion abnormalities by statistical comparison of atlas templates with images of patients. Methods of image registration of same modality and other image processing techniques were studied and used for comparison of patient images with the appropriate template. By the visual analysis of the templates it was found its validity as a representative image of normal perfusion. For the detection evaluation, the situation of myocardial segments (normal or abnormal) indicated by the detection algorithm was compared with the situation indicated in the medical appraisal report obtained by agreement of two specialists in order to determine the agreement and disagreement of the technique regarding the medical appraisal report and obtaining the statistical significance. Thus, there was a positive agreement index of the technique regarding the medical appraisal report of approximately 50%, a negative agreement index close to 82% and a general agreement index near 68%. The Fisher exact test was applied to the contingency tables, yielding a two-sided p-value less than 0.0001, that indicates a very low probability of the agreements have been obtained by chance. Algorithm improvements should be implemented and further tests with an effective gold-standard will be conducted to validate the technique.

O objetivo deste estudo foi comparar o eixo de rotação verdadeiro com o anatômico em ambiente virtual 3D, e seus efeitos sobre dois pontos anatômicos mandibulares. O eixo verdadeiro foi determinado em 14 indivíduos por meio de axiografia, e transferido para o ambiente virtual por TCFC, e posteriormente determinado anatomicamente, onde foram medidas as distâncias entre ambos. Foram simuladas rotações de 2º, 5º e 8º da mandíbula nos dois eixos, tanto para abertura como fechamento, e quantificadas as diferenças nos pontos da linha média inferior (LMI) e pogônio (Pg). O teste t pareado foi utilizado para examinar as diferenças entre as médias nas posições desses pontos (p<0,05). Os eixos verdadeiros localizaram-se dentro de um raio de 5 mm do anatômico em 67,86% da amostra. A distância absoluta média entre os eixos foi de 4,79 mm, enquanto que a vetorial foi de 2,33 no plano horizontal e 3,03 mm no vertical, resultando na direção anteroinferior em 71,43% dos eixos verdadeiros. Houve diferença estatisticamente significante na posição dos pontos LMI e Pg para todas as magnitudes e direções, entre os eixos. O eixo verdadeiro está localizado na direção anteroinferior em relação ao anatômico. Os efeitos na mandíbula são significantes e diferentes em todas as amplitudes, tanto para abertura como fechamento, porém com possível pequena relevância clínica.
The aim of this study was to compare the true hinge axis to the anatomic one in a virtual 3D environment, and also their respective effects on two mandibular anatomic points. The true axis has been determined in 14 individuals by means of axiography, and later transferred to a virtual environment by CBTC, where the anatomical axis was determined, and measured the distances between them. Mandibular rotation of 2º, 5º and 8º in both axes were performed, both for opening and closing, as well as the quantification of the difference found in the points of the lower midline (LM) and pogonion (Pg). Paired t-test was used to examine differences between the average values in the position of those points (p<0,05). The true axis was located within a 5mm-radius of the anatomic axis throughout 67.86% of the sample. The average absolute distance between the axes was 4.79 mm, while the vector distance was 2.33 mm in the horizontal plane e 3.03mm in the vertical plane, amounting to an anteriorinferior direction of 71.43% of the true axis. There was significant difference in the position of points LM and Pg to all magnitudes and directions within the axes. The true hinge axis is located in the anterior-inferior direction in relation to the anatomic axis. The effects observed onto the mandible are significant and different in all amplitudes, both for opening and closing positions, however they present small clinical relevance.

La maladie d’Alzheimer (MA) est la maladie neurodégénérative--incurable et irréversible pour le moment--la plus répandue chez les personnes âgées. On s’attend à ce qu’elle soit diagnostiquée à son stade précoce, Mild Cognitive Impairment (MCI), pour pouvoir intervenir et retarder son apparition. La tomographie par émission de positons au fluorodésoxyglucose (TEP-FDG) est considérée comme une modalité efficace pour diagnostiquer la MA et la phase précoce correspondante, car elle peut capturer les changements métaboliques dans le cerveau, indiquant ainsi des régions anormales. Cette thèse est consacrée à identifier et distinguer, sur des images TEP, les sujets atteints de MA de ceux qui sont sains. Ce travail vise également à prédire la conversion de MCI sous la modalité d’imagerie TEP-FDG. A cette fin, trois nouvelles méthodes indépendantes sont proposées.La première méthode est axée sur le développement de connectivités entre les régions anatomiques impliquées dans les images au TEP-FDG, qui sont rarement abordées dans les méthodes déjà publiées. Ces connectivités sont représentées par des similarités ou des mesures graphiques entre régions. Combinées ensuite aux propriétés de chaque région, ces caractéristiques sont intégrées dans un cadre de classification d’ensemble conçu pour résoudre les problèmes de diagnostic MA et de prédiction de conversion MCI.La seconde méthode étudie les caractéristiques permettant de caractériser les images au TEP-FDG à partir de gradients spatiaux, ce qui permet de lier les caractéristiques couramment utilisées, voxel ou régionales. Le gradient spatial est quantifié par un histogramme 2D d’orientation et exprimé sous forme multi-échelle. Les résultats sont obtenus en intégrant différentes échelles de gradients spatiaux dans différentes régions.La troisième méthode applique le Convolutional Neural Network (CNN) sur les trois axes des données 3D de TEP-FDG, proposant ainsi la principale architecture CNN à vues multiples. Une telle architecture peut faciliter les opérations de convolution, de la 3D à la 2D, tout en tenant compte des relations spatiales, qui bénéficient d’une nouvelle couche de cartographie. Ensuite, le traitement sur les trois axes sont combinées et prennent une décision conjointement.Les expériences menées sur des ensembles de données publics montrent que les trois méthodes proposées peuvent atteindre des performances significatives et, de surcroît, dépasser les approches les plus avancées
Alzheimer's disease (AD) is becoming the dominant type of neurodegenerative brain disease in elderly people, which is incurable and irreversible for now. It is expected to diagnose its early stage, Mild Cognitive Impairment (MCI), then interventions can be applied to delay the onset. Fluorodeoxyglucose positron emission tomography (FDG-PET) is considered as a significant and effective modality to diagnose AD and the corresponding early phase since it can capture metabolic changes in the brain thereby indicating abnormal regions. Therefore, this thesis is devoted to identify AD from Normal Control (NC) and predict MCI conversion under FDG-PET modality. For this purpose, three independent novel methods are proposed. The first method focuses on developing connectivities among anatomical regions involved in FDG-PET images which are rarely addressed in previous methods. Such connectivities are represented by either similarities or graph measures among regions. Then combined with each region's properties, these features are fed into a designed ensemble classification framework to tackle problems of AD diagnosis and MCI conversion prediction. The second method investigates features to characterize FDG-PET images from the view of spatial gradients, which can link the commonly used features, voxel-wise and region-wise features. The spatial gradient is quantified by a 2D histogram of orientation and expressed in a multiscale manner. The results are given by integrating different scales of spatial gradients within different regions. The third method applies Convolutional Neural Network (CNN) techniques to three views of FDG-PET data, thereby designing the main multiview CNN architecture. Such an architecture can facilitate convolutional operations, from 3D to 2D, and meanwhile consider spatial relations, which is benefited from a novel mapping layer with cuboid convolution kernels. Then three views are combined and make a decision jointly. Experiments conducted on public dataset show that the three proposed methods can achieve significant performance and moreover, outperform most state-of-the-art approaches

Made available in DSpace on 2016-08-17T14:53:07Z (GMT). No. of bitstreams: 1 Stelmo Magalhaes Barros Netto.pdf: 2768924 bytes, checksum: bf6f24780a03adb4f2940b818c95f293 (MD5) Previous issue date: 2010-02-10
Conselho Nacional de Desenvolvimento Científico e Tecnológico
Lung cancer is still one of the most frequent types throughout the world. Its diagnosis is very difficult because its initial morphological characteristics are not well defined, and also because of its location in relation to the lung. It is usually detected late, fact that causes a large lethality rate. Facing these difficulties, many researches are done, concerning both detection and diagnosis. The objective of this work is to propose a methodology for computer-aided automatic lung nodule detection. The return of the development of such methodology is that its application will aid the doctor in the simultaneous detection of several nodules present in computerized tomography images. The methodology developed for automatic detection of lung nodules involves the use of a method of competitive learning, called Growing Neural Gas (GNG). The methodology still consists in the reduction of the volume of interest, by the use of techniques largely used in thorax extraction, lung extraction and reconstruction. The next stage is the application of the GNG in the resulting volume of interest, that together with the separation of the nodules from the various structures present in the lung form the segmentation stage, and, finally, through texture and geometry measurements, the classification as either nodule or non-nodule is performed. The methodology guarantees that nodules of reasonable size are found with sensibility of 86%, specificity of 91%, what results in accuracy of 91%, in average, for ten training and test experiments, in a sample of 48 nodules occurring in 29 exams. The false-positive per exam rate was of 0.138, for the 29 analyzed exams.
O câncer de pulmão ainda é um dos mais incidentes em todo mundo. Seu diagnóstico é de difícil realização, devido as suas características morfológicas iniciais não estarem bem definidas e também por causa da sua localização em relação ao pulmão. É geralmente detectado tardiamente, que tem como conseqüência uma alta taxa de letalidade. Diante destas dificuldades muitas pesquisas são realizadas, tanto em relação a sua detecção, quanto a seu diagnóstico. O objetivo deste trabalho é propor uma metodologia de detecção automática do nódulo pulmonar com o auxílio do computador. O ganho com o desenvolvimento desta metodologia, é que sua implementação auxiliará ao médico na detecção simultânea dos diversos nódulos presentes nas imagens de tomografia computadorizada. A metodologia de detecção de nódulos pulmonares desenvolvida envolve a utilização de um método da aprendizagem competitiva, chamado de Growing Neural Gas (GNG). A metodologia ainda consiste na redução do volume de interesse, através de técnicas amplamente utilizadas na extração do tórax, extração do pulmão e reconstrução. A etapa seguinte é a aplicação do GNG no volume de interesse resultante, que em conjunto com a separação do nódulo das diversas estruturas presentes formam a etapa de segmentação, e por fim, é realizada a classificação das estruturas em nódulo e não-nódulo, por meio das medidas de geometria e textura. A metodologia garante que nódulos com tamanho razoável sejam encontrados com sensibilidade de 86%, especificidade de 91%, que resulta em uma acurácia de 91%, em média, para dez experimentos de treino e teste, em uma amostra de 48 nódulos ocorridos em 29 exames. A taxa de falsos positivos por exame foi de 0,138, para os 29 exames analisados.

3D printing applications have achieved increased success as an additive manufacturing (AM) process. Micro-structure of mechanical/biological materials present design challenges owing to the resolution of 3D printers and material properties/composition. Biological materials are complex in structure and composition. Efforts have been made by 3D printer manufacturers to provide materials with varying physical, mechanical and chemical properties, to handle simple to complex applications. As 3D printing is finding more medical applications, we expect future uses in areas such as hip replacement - where smoothness of the femoral head is important to reduce friction that can cause a lot of pain to a patient. The issue of print resolution plays a vital role due to staircase effect. In some practical applications where 3D printing is intended to produce replacement parts with joints with movable parts, low resolution printing results in fused joints when the joint clearance is intended to be very small. Various 3D printers are capable of print resolutions of up to 600dpi (dots per inch) as quoted in their datasheets. Although the above quoted level of detail can satisfy the micro-structure needs of a large set of biological/mechanical models under investigation, it is important to include the ability of a 3D slicing application to check that the printer can properly produce the feature with the smallest detail in a model. A way to perform this check would be the physical measurement of printed parts and comparison to expected results. Our work includes a method for using ray casting to detect features in the 3D CAD models whose sizes are below the minimum allowed by the printer resolution. The resolution validation method is tested using a few simple and complex 3D models. Our proposed method serves two purposes: (a) to assist CAD model designers in developing models whose printability is assured. This is achieved by warning or preventing the designer when they are about to perform shape operations that will lead to regions/features with sizes lower than that of the printer resolution; (b) to validate slicing outputs before generation of G-Codes to identify regions/features with sizes lower than the printer resolution.

Fluorescence microscopy is an essential tool in biomedical sciences that allows specific molecules to be visualized in the complex and crowded environment of cells. The continuous introduction of new imaging techniques makes microscopes more powerful and versatile, but there is more than meets the eye. In addition to develop- ing new methods, we can work towards getting the most out of existing data and technologies. By harnessing unused potential, this work aims to increase the richness, reliability, and power of fluorescence microscopy data in three key ways: through standardization, evaluation and innovation. A universal standard makes it easier to assess, compare and analyze imaging data – from the level of a single laboratory to the broader life sciences community. We propose a data-standard for fluorescence microscopy that can increase the confidence in experimental results, facilitate the exchange of data, and maximize compatibility with current and future data analysis techniques. Cutting-edge imaging technologies often rely on sophisticated hardware and multi-layered algorithms for reconstruction and analysis. Consequently, the trustworthiness of new methods can be difficult to assess. To evaluate the reliability and limitations of complex methods, quantitative analyses – such as the one present here for the 3D SPEED method – are paramount. The limited resolution of optical microscopes prevents direct observation of macro- molecules like DNA and RNA. We present a multi-color, achromatic, cryogenic fluorescence microscope that has the potential to produce multi-color images with sub-nanometer precision. This innovation would move fluorescence imaging beyond the limitations of optics and into the world of molecular resolution.

博士
國立清華大學
生醫工程與環境科學系
103
Computed tomography (CT) can rapidly provide high resolution cross-section images. It has become one of the powerful tools in clinical and has been widely applied to achieve a variety of diagnostic and therapeutic purposes. In recent years, number of CT procedures increase year after year with an annual growth rate higher than 10% in Taiwan and in the U.S. Comparing to the other radiological examinations, CT scan delivers relatively high radiation dose to patients. From the view point of radiation protection, medical exposure is justified as long as it follows the ALARA principle. Currently, CT images are mainly used in visually diagnosis of various diseases but have no other way to effective utilize. In addition to the visual observation, tissue parameters, such as physical electron densities, effective atomic numbers and bone mineral densities, obtained from CT image-based quantification are also useful for several physical correction and diagnosis in clinical. However, common CT scanners employ polychromatic X-ray spectrum and cumulative detector, causes the composition and attenuation information of scanning objects are difficult to estimate from acquired projection data or reconstructed images. Therefore, present CT image-based quantification is mainly performed through various tissue equivalent materials (TEMs). Nevertheless, the differences between the elemental composition of tissue equivalent materials and actual human tissues results that the estimated parameters are a reference equivalent. In view of the above, a novel computer-aided quantification (CAQ) method was proposed to achieve fast and accurate tissue parameter quantification. In this method, a stoichiometric calibration was performed to acquire spectrum characteristic parameters (SCPs) that describe the energy spectrum of a specific CT scanner. The acquired SCPs were then used to convert CT number into clinically valuable physical and physiological tissue parameters (PTPs and PoTPs). This study was divided in to two parts. In the first part, the CT number was converted into the PTPs by using conversion relationships. In addition, these parameters were further used to calculate the mass attenuation coefficients (MACs) and mass energy transfer coefficients (MEACs) with physical models. In the second part, the CT numbers were converted into bone physiological parameters (BPPs) by using a novel mixture model. The results show that the proposed CAQ method can accurately convert the CT images into PTP and BPP maps. Moreover, the proposed method also reduce the influences of energy spectrum that is helpful in image exchanging and comparing between scanners. We conclude that the proposed CAQ method could be applied in the clinical to estimate several tissue parameters from CT image for various diagnostic and therapeutic purposes, whereby benefits for patients from CT examinations can be increased.

碩士
中國醫藥大學
臨床醫學研究所碩士班
100
Hepatic Carcinoma is the most common cancer disease among Taiwan population. With a good and healthy daily life planning, having a regularly health exam can also keep the disease away. Generally speaking, Computer Tomography is the most commonly used instrument in the diagnosis, and the most important part is to define the location and contour of liver tumor, in order to assist the radiologists in diagnosis and pretreatment evaluation. In this study, we develop a semi-automatic computer-aided detection system, which is able to achieve an objective and consistent result in diagnosis and treatment of liver tumor. According to the experiment, the relative accuracy and error rate are 85.6% and 14.4%.

碩士
國立臺灣大學
光電工程學研究所
106
Optical coherence tomography (OCT) is an important three-dimensional optical imaging technique in biomedical realm. It has non-invasive, label-free, and high spatial resolution characteristics. Based on its unique advantages, OCT plays a crucial role in clinical diagnosis in ophthalmology and cardiovascular system. However, in dermatology and gastroenterology, the clinical diagnosis for skin cancer and colon cancer still relies on biopsy, which is an invasive procedure. Biopsy in skin will cause bleeding and may leave scars. For diagnosis of colon cancer, increasing the amount of biopsy will increase the chance of bleading and even the possibility of perforation of the large intestine. Besides, whether it is skin cancer or colon cancer, doctors need to rely on biopsy to check whether the tumor area is removed thoroughly, which means that patients may need second or multiple operations. The Mirau-based OCT system used in this thesis has high lateral resolution of 0.8 μm and axial resolution of 0.9 μm. A homemade 〖Ce〗^(3+):YAG crystal fiber spontaneous emission (SE) light source was used to build the FF-OCT system. The experiments were conducted in two separate parts: melanoma part and colon cancer part. In the melanoma part, FF-OCT with a XY stage was used to scan 24 sets blank tissue sections. A set includes OCT images of normal skin tissue and melanoma tissue. Based on the ground truth provided by a dermatologist, regions were chosen to be training set and testing set. Nineteen features of lateral and vertical were extracted from these regions. The discriminant model is established by 19 features extracted from the training set, and the discriminant model was applied to the testing set to distinguish normal tissue and melanoma tissue. The lateral discriminant resolution is 48 μm*54 μm (108 pixel *122 pixel), and the discriminant algorithm is linear discriminant analysis (LDA). The mean discriminant accuracy in 24 sets OCT images is 87.5%. In the colon cancer part, FF-OCT with a XY stage was used to scan 16 sets unstained tissue sections which include normal large intestine tissue and colon cancer tissue. Based on the ground truth provided by a pathologist, 22 features were extracted from regions including tissues in OCT images. Based on the discriminant model established by 22 features extracted from the training set, the mean discriminant accuracy in the 16 sets OCT images is 87.4%. The lateral discriminant resolution is 222 μm*222 μm (500 pixel *500 pixel), and the discriminant method is support vector machine (SVM). The best discriminant resolution of colon cancer is much larger than Melanoma. It is related to the characteristics of the extracted lateral features and the periodic structure of large intestine tissues. This thesis presents one algorithm for discrimination of OCT images of melanoma and normal skin tissue, another for discrimination of OCT images of colon cancer and normal large intestine tissue. This study provides a preliminary study for the potentially possible applications of OCT in skin cancer and colon cancer diagnosis in vivo.

碩士
慈濟大學
醫學資訊學系碩士班
99
Bone metastases are commonly diagnosed in patients with advanced breast cancer, especially in vertebrae. Bone metastases can appear lytic, blastic, or anywhere in a continuum between these extremes. The presence or absence of bone metastases is a critical issue in the initial staging and follow-up of breast cancer because it can directly alter the therapeutic strategy. In this paper, we aim to developing a computer-aided diagnosis system for the detection of metastasis in vertebrae at whole body CT. We developed an automated method to extract ROIs of trabecular centrum from vertebrae. We computed 11 texture features and their inter-slice differences for each ROI. Total 33 features were fed into an MLP neural network to identify whether there is any abnormality in the trabecular centrum. The average accuracy, sensitivity, and specificity were 89.73%, 85.41% and 91.84%, respectively. The FN was substantially decreased from 20.83% to 14.58% when the inter-slice difference features were used.

Research Doctorate - Doctor of Philosophy (PhD)
The liver is one of the most important organs in the human body. It carries out a variety of functions including filtering the blood, making bile and proteins, processing sugar, breaking down medications, and storing iron, minerals, and vitamins. However, the liver is prone to many diseases such as hepatitis C, cirrhosis, and cancer. As computer science and technology advances, computer-aided surgical planning systems have played an important role in the diagnosis and treatment of liver diseases. These systems can present the structures of various liver vessels, generate resection proposals, offer 3D visualizations, provide surgical cutting simulations, and shorter planning times. However, among these systems, one of the most challenging issues is the accurate segmentation of the liver from its surrounding organs in computed tomography images. Factors contributing to the challenge in carrying out accurate liver segmentation include the similar intensity values between adjacent organs, geometrically complex liver structure, and the injection of contrast media that causes all tissues to have similar gray-level values. Several artefacts of pulsation and motion, and partial volume effects, also increase difficulties for automatic liver segmentation in computed tomography images. Significant individual variations in shape and volume of the liver also add to the difficulties. Therefore, liver segmentation from medical images remains an open problem.In this research, we aim to perform accurate and automatic 3D liver segmentation from the latest multi-slice spiral/helical computed tomography (CT) scans, an achievement which would be very useful for computer-aided surgical planning systems. The development and evaluation of a clinically applicable segmentation algorithm, and its integration into software that could be used by medical experts, represents the major focus of the thesis. Level set methods have been widely used in medical image segmentation and perform well in segmenting irregularly shaped objects such as the liver. However, level set methods fail to segment meaningful objects from images if the objects are occluded by other objects, if some parts have low contrast (or are even missing), or if the target object has similar intensity values to adjacent objects. Since all these factors exist in the case of the liver, classical level set methods are not well suited to accurately segment the liver from abdominal CT scans. In this thesis, the enhanced level set method has been modified to make it suitable for segmenting the liver from an abdominal CT scan. We have improved the level set method to enable segmentation of the liver parenchyma from CT images by introducing a priori knowledge about the liver into the level set framework. These improvements make it possible to distinguish unclear liver boundaries, prevent surrounding organs from confusing the boundaries, and enhance segmentation performance. An important aspect of our improvements is that implementation of the necessary prior knowledge is not long or difficult compared to other segmentation methods. In initial exploratory work, the novel liver segmentation algorithm we first developed used the level set method together with an intensity prior (IP). The IP model improved the level set method by adding a priori statistical knowledge about the intensity distribution inside and outside the liver to the level set framework. The main merits of this approach were found to be its strong ability to dynamically guide the direction of the evolving contour and prevent it from leaking into regions with unclear boundaries. Examples of applying the proposed IP algorithm on real computed tomography images are presented. We show that the proposed method can deliver superior segmentation compared to the distance-regularized level set (DRLS) method. The average accuracy values for the IP model and the DRLS model are 99% and 89%, respectively. However, the IP model does have some limitations. We need to train the algorithm on liver slices that have a very similar intensity distribution to the target. This indicates that the statistical learning applied a priori in the training stage cannot be generally transferred to a large range of liver slices. Consequently, the method is not capable of segmenting a sequence of liver slices and building a complete liver volume. This motivated us to develop a liver segmentation algorithm which used the level set method together with density matching and a shape prior (DMSP). The DMSP model we developed provides accurate and automatic 3D liver segmentation from abdominal CT images. The algorithm is novel in that it combines density matching with prior knowledge about the liver shape. Density matching is a tracking method which maximizes the Bhattacharyya similarity measure between the photometric distribution inside the evolving curve and a model photometric distribution learned a priori. Density matching provides adaptive shrinkage or expansion to the evolving contour, while the shape prior improves robustness of the density matching and discourages the evolving contour from exceeding liver boundaries at regions which are unclear. For the purpose of comparison, we improved the IP model by adding a shape prior to its framework, producing an intensity prior and shape prior model, or IPSP model. However, even with this modification, the learning of the a priori statistical model applied during the training stage could still not correctly allow a liver volume to be reconstructed from a sequence of liver slices. Comparison experiments have shown that the DMSP model outperformed the IPSP model and performed well for all the investigated liver cases in our test data. The average overlap values for the IPSP model and the DMSP model were 76% and 91%, respectively. We compared the DMSP model with several other reported methods: the density matching (DM) model, the overlap prior (OP) model, and the DRLS method. Comparisons showed that the proposed method achieved better performance than any of these aforementioned approaches. The proposed method was shown to be more effective in overcoming over- and under-segmentation problems. The average overlap values of segmentation (compared to the ground truth) were estimated to be 69%, 77%, 63%, and 93% for the DM model, OP model, DRLS model, and DMSP model, respectively. Since the DMSP model achieves better performance than previous analogous studies, it has the potential to be used in clinical practice or in a computer-aided surgical planning system.

博士
義守大學
資訊工程學系
104
In recent years, whole body bone scan imaging (WBBS) has become an important and widespread diagnostic tool in nuclear medicine due to its high sensitivity and relatively low cost. WBBS is particularly important because it can identify bony metastasis; however, it is limited in some cases; for example, osteolytic bony lesions. Additional factors, such as patients’ individual differences, poor image quality, and doctor experiences, can bias the interpretation of WBBS and affect the accuracy of diagnosis and treatment. Therefore, the development of a computer-aided diagnosis (CAD) system to provide objective and quantitative analysis for WBBS is an important clinical research issue. In our study, we developed an automated detection system – the abnormal flow browser irregular flux viewer (IFV), with the ability to automatically locate abnormal flow in bony lesions; this tool could provide assistance to physician diagnosis and give a prediction value in bone metastasis. The system was developed in two stages. In the first stage, we tried to perform “non-supervision type of neural network training” to find the gradient and kinetic energy of the index value. Bone scan images of three types of cancer patients (prostate, lung, and breast cancers) were collected. The bone scan results were categorized into four groups (No Metastasis, degenerative arthritis, slight bony metastasis, or serious bony metastasis). Using Gradient Vector Flow, we assessed different areas of bone image pixels to calculate the values of gradient and momentum for adaptive threshold.　In the second stage, we used View-Tool (an abnormal flow browser for assessing the abnormal flow point of the clustering analysis) to correct the image histogram in order to obtain “self-cluster” and “union-cluster” indexes, according to the correlated distance from the centroid to distinguish abnormal flow accumulated points (hot spots). Then, the hot spots of the pixels were labeled as the suspected lesions. We tried to compare the clinical diagnostic reports of CT, MRI, SPECT/CT, and PET/CT with IFV-BS reports. Our proposed approach had a higher sensitivity to improve the inherent limits of osteolytic lesion in planar bone scintigraphy. The corresponding results show sensitivity to predict skeletal metastasis in prostate cancer (93 %) [95 % confidence interval (CI) 0.91~0.93], breast cancer (91 %) [95 % CI 0.90~0.92], and lung cancer (83 %) [95 % CI 0.82~0.84]. The results of our study showed that our abnormal flow browser is reliable and may provide assistance for image interpretation and generate prediction values in WBBS.

博士
臺灣大學
醫學工程學研究所
98
Intracranial hematomas, either traumatic or spontaneous, can produce fatal outcomes because they can produce local pressure on the brain. Accurate diagnosis and rapid decision making are the key factors to good patient outcome. This thesis introduces new methods capable of obtaining the features of the intracranial hematomas in brain CT images. In addition, a new approach of image segmentation integrating binary level set method and multi-resolution processing is proposed. We develop the decision rules to recognize the type of the intracranial hematoma on CT slices with large intracranial hematomas using C4.5 algorithm. These decision rules work well in different resolutions. To obtain robust segmentation of the intracranial hematoma regions, we introduce a multi-resolution binary level set method using image pyramids and apply it to hematoma segmentation. Prior to segmentation of the hematoma from the brain, anatomical knowledge is integrated with image processing techniques in the segmentation of intracranial regions. The results show excellent precision and recall as verified by human experts. In the second half of this thesis, we describe two methods for automatic measurement of the midline shift (MLS). The first one employs symmetry and curve fitting to measure the MLS of the CT slice at the level of Foramen of Monro. Genetic algorithm is used for parameter optimization. Landmark-based MLS recognition is carried out by first segmenting the frontal horn region followed by a knowledge-driven rule. Hough transform (HT) is then applied to locate the septum pellucidum. Finally, we describe automatic recognition of the basal cisterns using HT. This method is able to pick out the normal or compressed basal cistern region from the given CT data set.