Дисертації з теми "Glaucome – Diagnostic"

Les pathologies oculaires sont au cœur d'enjeux majeurs de santé publique. L'une d’elles, le glaucome, nécessite un dépistage précoce pour garantir le traitement efficace des patients atteints, et prévenir de déficiences visuelles irréversibles. L'avènement des approches dites de vision par ordinateur et d'apprentissage profond est à l'origine d'un changement de paradigme dans le domaine de l'ophtalmologie, vecteurs d'appui sans précédent dans les choix diagnostiques et thérapeutiques. Dans cette thèse, nous proposons de nouvelles méthodes pour le développement de systèmes intelligents dédiés au dépistage précoce de glaucome sur des images de la rétine. Nous visons notamment le déploiement de systèmes mobiles intégrant des dispositifs d'acquisition portables, pour le dépistage de proximité et/ou itinérant.Dans un premier temps, nous proposons une méthode permettant l'analyse de la tête du nerf optique (communément appelée papille), région caractérisée par un changement topographique en présence de glaucome. S’appuyant sur un algorithme précis de segmentation des structures de la papille, notamment du disque optique et de l’excavation (cup) en son sein, la méthode extrait des mesures cliniques pertinentes comme le ratio cup-disque, les secteurs inférieur-supérieur-nasal-temporal (ISNT) et l'aire de l'anneau neuro-rétinien. Un protocole clinique basé sur des références en ophtalmologie permet alors de dépister les cas glaucomateux, et donner des indications sur les stades de développement de la neuropathie (glaucome précoce, modéré ou avancé). Bien que très précise dans le dépistage, avec un taux de performance de 94% sur la base d'évaluation (DRISHTI-GS1), cette méthode a permis de mettre en lumière la nécessité d’améliorer la capacité de généralisation de l’outil, notamment en présence de papilles glaucomateuses sans excavation (faux négatifs) ou de larges papilles saines (faux positifs). Dans un deuxième temps, nous proposons une méthode basée sur des algorithmes d’apprentissage profond, permettant une interprétation automatique des traits caractéristiques de rétines saines ou glaucomateuses. Ce travail exploite des réseaux de neurones convolutionnels de l’état de l’art (VGG-16, ResNet50, Inception-v3, MobileNet et DenseNet121), et propose une méthode d'apprentissage par transfert efficace pour adapter ces réseaux à la tâche de dépistage de la pathologie. Ces modèles atteignent une AUC de plus de 0.97, toutefois, cette étude comparative a permis de déceler les besoins nécessaires au développement de modèles performants, en vue d'un déploiement en condition clinique : 1) une base d’images rétiniennes consistante en termes de taille, d'équilibre inter-classe, de fiabilité diagnostique et de variabilité clinique, 2) des modèles interprétables et explicables, permettant aux spécialistes de comprendre et discuter le résultat de dépistage. Dans ce sens, nous proposons dans un troisième temps une méthode exploitant les percées récentes de l’apprentissage semi-supervisé, pour la génération d’images rétiniennes synthétiques. L’algorithme proposé, BAGAN (pour Balancing GAN), permet de produire à partir d’une base d’images de référence (REFUGE), une base rétinienne comblant le déséquilibre inter-classes potentiellement responsable de biais diagnostiques, tout en répondant aux critères de qualité d’image et de diversité clinique. Nous avons démontré la pertinence d’une telle base dans le développement d’algorithmes semi-supervisés pour le dépistage du glaucome.Enfin, une toute nouvelle interface, disponible sur plateforme de bureau et mobile, a été conçue à destination des ophtalmologistes et professionnels de santé. Ludique et intuitive, elle intègre différentes fonctionnalités basées sur les algorithmes développées, et permet un dépistage en temps réel pour contribuer à l'amélioration de la prise en charge en santé oculaire
Ocular diseases are at the core of major public health issues. One of them, glaucoma, requires early screening to ensure effective treatment of affected patients, and prevent irreversible visual damages. The advent of so-called computer vision and deep learning approaches has led to a paradigm shift in the field of ophthalmology, providing unprecedented support in diagnostic and therapeutic choices. In this thesis, we propose new methods for the development of intelligent systems dedicated to the early detection of glaucoma from retinal images. In particular, we aim at deploying of mobile-based computer-aided diagnosis systems, for remote screening. Firstly, we proposed a method aiming at analyzing the optic nerve head, featured by morphological changes in the presence of glaucoma. Based on a precise algorithm for segmenting the structures of the optic disc and the cup within it, the method extracts clinically relevant measures such as the cup-to-disc ratio, the inferior-superior-nasal-temporal (ISNT) sectors and the neuroretinal rim area. A clinical protocol based on ophthalmic references is drawn to screen for glaucoma, and give indications about the stages of development of the neuropathy (early, moderate or advanced glaucoma). Although very accurate screening, with a performance rate of 94% on the evaluation base (DRISHTI-GS1), this method has highlighted the need to improve generalizability, particularly in the presence of glaucomatous nerve heads without excavation (false negatives) or large healthy nerve heads (false positives). Secondly, we proposed a method based on deep learning algorithms, allowing an automated interpretation of healthy or glaucomatous retinas. This work exploits state-of-the-art convolutional neural networks (VGG-16, ResNet50, Inception-v3, MobileNet and DenseNet121), and proposes an efficient transfer learning method to adapt these networks to the glaucoma screening. These models achieve an AUC of more than 0.97, however, this comparative study has identified the needs for developing efficient models for deployment in clinical conditions: 1) a consistent retinal image dataset in terms of size, inter-class balance, diagnostic reliability and clinical variability, 2) interpretable and explainable models, allowing specialists to understand and discuss the screening result.In this sense, we propose in a third step a method exploiting recent advances in semi-supervised learning, for the generation of synthetic retinal images. The proposed algorithm, BAGAN (for Balancing GAN), allows to produce from a reference image dataset (REFUGE), a new dataset filling the inter-class imbalance potentially responsible for diagnostic bias, while meeting the criteria of image quality and clinical diversity. We have demonstrated the relevance of such dataset in the further development of semi-supervised diagnosis algorithms. Finally, a brand new interface, available on desktop and mobile platforms, has been designed for ophthalmologists and health professionals. Smart and intuitive, it integrates various functionalities based on the developed algorithms, and allows real-time screening to contribute to the improvement of eye health care

Submitted by Rosivalda Pereira (mrs.pereira@ufma.br) on 2017-05-30T20:52:23Z No. of bitstreams: 1 JeffersonSousa.pdf: 2049556 bytes, checksum: f9f7046416ba8b0f024d06906537417f (MD5)
Made available in DSpace on 2017-05-30T20:52:23Z (GMT). No. of bitstreams: 1 JeffersonSousa.pdf: 2049556 bytes, checksum: f9f7046416ba8b0f024d06906537417f (MD5) Previous issue date: 2017-01-26
Glaucoma is one of the diseases that mopstly causes blindness, according to estimates by the World Health Organization (WHO). The Brazilian Council of Ophthalmology (CBO) estimates that in Brazil there are 985 thousand glaucoma patients with more than 40 years of age. The use of computer aided detection and diagnosis systems (CAD - Computer Aided Detection and CADx - Computer Aided Diagnosis) has contributed to increase the chances of detection and correct diagnoses. They provide a second opinion, assisting the experts in making decisions about the treatment of glaucoma. The main objective of this work is to present a method for automatic diagnosis of glaucoma in retinography images using texture characteristics. The Local Binary Pattern is used to generate a representation of the texture patterns of the image. The geostatistical functions, semivariogram, semimagram, covariogram and correlogram, are used as texture extractors. With the generated characteristics, a step is made to select the best classification model using the genetic algorithm. Then sorting is performed using the Support Vector Machine. The best result was an accuracy of 91%, sensitivity of 95% and specificity of 88%, proving that the characteristics generated by the geostatistical functions for texture extraction generate a satisfactory discriminant set.
O glaucoma e uma das doenças que mais causam cegueira em todo o mundo segundo estimativa da Organização Mundial da Saúde (OMS). O Conselho Brasileiro de Oftalmologia (CBO) estima que no Brasil existam 985 mil portadores de glaucoma com mais de 40 anos de idade. A utilização de sistemas de detecção e diagnóstico auxiliados por computador (CAD - Computer Aided Detection e CADx -Computer Aided Diagnosis) tem contribuído para aumentar as chances de detecção e diagnósticos corretos, que proporcionam, uma opinião preliminar, auxiliando os especialistas na tomada de decisões sobre o tratamento do glaucoma. O principal objetivo deste trabalho e apresentar um método para diagnóstico automático de glaucoma em imagens de retinografia utilizando características de textura. Neste trabalho, o Local Binary Pattern é usado para gerar uma representação do padrões de textura da imagem. As funções geoestatísticas, semivariograma, semimadograma, covariograma e correlograma, sao utilizadas como extratores de textura. Com as características geradas e realizada uma etapa para selecionar o melhor modelo de classificação utilizando o algoritmo genético. Em seguida é realizada a classificação usando a Maquina de Vetores de Suporte. O método apresentou como melhor resultado uma acurácia de 91%, sensibilidade de 95% e especificidade de 88%, comprovando que as características geradas pelas funções geoestatísticas para extração de textura geram um conjunto discriminante satisfatório.

L'apprentissage multi-tâches est un paradigme d'apprentissage impliquant l’optimisation de paramètres par rapport à plusieurs tâches simultanément. En apprenant plusieurs tâches liées, un modèle d'apprentissage dispose d'un ensemble d'informations plus complet concernant le domaine dont les tâches sont issues, lui permettant ainsi de construire un meilleur ensemble d’hypothèse sur ce domaine. Cependant, en pratique, les gains de performance obtenus par les réseaux multi-tâches sont loin d'être systématiques. Il arrive au contraire que ces réseaux subissent une perte de performance liée à des phénomènes d’interférences entre les différentes tâches. Cette thèse traite du problème d'interférences en apprentissage multi-tâches, afin d'améliorer les capacités de généralisation des réseaux de neurones profonds
Multi-task learning (MTL) is a learning paradigm involving the joint optimization of parameters with respect to multiple tasks. By learning multiple related tasks, a learner receives more complete and complementary information on the input domain from which the tasks are issued. This allows to gain better understanding of the domain by building a more accurate set of assumptions of it. However, in practice, the broader use of MTL is hindered by the lack of consistent performance gains observed by deep multi-task networks. It is often the case that deep MTL networks suffer from performance degradation caused by task interference. This thesis addresses the problem of task interference in Multi-Task learning, in order to improve the generalization capabilities of deep neural networks

This thesis describes a series of investigations into the use of optic nerve head (ONH) imaging in primary open-angle glaucoma (POAG), and its relation to visual function. Accurate diagnosis is a key issue in POAG, particularly the difficult task of separating those with early disease from those healthy individuals who display signs of POAG. The purpose of this work is to improve diagnostic methods in glaucoma through use of ONH imaging and its relationship with visual field (VF) loss. First, the performance of a group of expert clinicians evaluating ONH photographs for glaucomatous damage was investigated. The results showed that even when their assessments are combined discrimination between eyes with and without POAG (based on VF loss) is far from perfect, highlighting the need for improvements in diagnosis. The possibility of combining structural and functional data to aid diagnosis was then considered. This requires VF loss and ONH damage to be strongly topographically related. The strength of this relationship was evaluated in 185 patients with POAG. 10,000 computer-generated maps between the ONH and VF were tested and the topographic relationship measured with each of these was compared to that using a published structure-based map. The weak topographic relationships found suggest that the application of these maps to individual patients is limited with current measures. The next chapter describes how a multispectral imaging (MSI, also called hyperspectral imaging) system was set-up for spatial evaluation of ONH oxygenation using a Beer-Lambert law model. Test-retest repeatability was tested and found to be acceptable for the purposes of the following studies. The MSI system was then used for an investigation of the relationship between ONH oxygenation and VF loss. 33 eyes of 18 patients underwent VF testing, MSI and HRT3 imaging. Superior-inferior asymmetries in VF sensitivity were compared to superior-inferior asymmetries in ONH oxygenation measured by MSI and in neuroretinal rim (NRR) area measured by HRT3. This way we take advantage of the typical progression of POAG and each eye acts as its own reference, negating the effect of a wide normal range and overlap between health and disease. This study found, for the first time, a strong association between ONH oxygenation and VF sensitivity. A re-analysis of the 33 ONH oxygenation maps was then performed to assess oxygenation only in the area of the NRR as defined by the HRT. Superior-inferior asymmetries in NRR oxygenation were then compared to superior-inferior asymmetries in VF loss, and the associations found were similarly strong. This study shows that MSI is capable of detecting areas of NRR deemed healthy tissue by structural imaging techniques, which are in fact poorly oxygenated and associated with VF defects. These findings show that NRR oxygenation measured by MSI is strongly related to VF loss. This important information complements existing technologies and may aid in the future diagnosis and management of patients with POAG.

Orientadores: Vital Paulino Costa, Ana Maria Marcondes
Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Ciencias Medicas
Made available in DSpace on 2018-08-04T01:23:28Z (GMT). No. of bitstreams: 1 Silva_LeopoldoMagachodosSantos_D.pdf: 1066525 bytes, checksum: d1d302bd6421f23d5f4e474e23049056 (MD5) Previous issue date: 2004
Resumo: Objetivo: Avaliar a Sensibilidade e Especificidade de parâmetros estruturais (topografia de disco óptico e análise da camada de fibras nervosas da retina) no diagnóstico do glaucoma. Métodos: Pacientes com glaucoma primário de ângulo aberto foram selecionados do ambulatório de glaucoma da Unicamp. Indivíduos normais foram recrutados junto a voluntários dentro da equipe médica, amigos e parentes dos pacientes. Os participantes, após aprovação do Comitê de Ética da instituição, foram submetidos a exame oftalmológico completo, campimetria computadorizada (Humphrey Full Threshold, programa 24-2), topografia de disco óptico com o TOPSS e análise da camada de fibras nervosas da retina por meio da polarimetria de varredura a laser (GDx, versão 1.0.12). Curvas ROC foram criadas para se estabelecer novos pontos de corte para cada parâmetro individual de ambos os instrumentos. Em seguida, a análise de regressão multivariada foi empregada com parâmetros da topografia de disco óptico e polarimetria de varredura a laser separados e em conjunto para obter uma melhor relação sensibilidade/especificidade (Se/Es) no diagnóstico do glaucoma. Resultados: Cento e doze indivíduos com glaucoma e 88 normais foram incluídos. Os parâmetros individuais da polarimetria de varredura a laser com melhor capacidade em diferenciar olhos normais de glaucomatosos foram: The Number (Se: 79,5%, Es: 81,8%, área abaixo da curva ROC - aROC: 0,870), Maximum Modulation (Se: 83,0%, Es: 76,1%, aROC: 0,842) e Ellipse Modulation (Se: 65,2%, Es: 88,6%, aROC: 0,831), enquanto Average Disc Diameter (Se: 64%, Sp: 89%, aROC: 0.824), Total Disc Area (Se: 85%, Sp: 66%, aROC: 0.802) e Cup Area (Se: 69%, Sp: 85%, aROC: 0.797) foram os melhores parâmetros individuais da topografia de disco óptico. Entretanto, a análise discriminante apresentou melhores resultados que os parâmetros individuais, tanto para a polarimetria de varredura a laser (Se: 87,5%, Es: 86,4%, aROC: 0,900), topografia de disco óptico (Se: 90%, Sp: 81%, aROC: 0,910), quanto para a combinação de parâmetros derivados de ambos os instrumentos (Se: 93%, Es: 91%, aROC: 0,970). Conclusão: Parâmetros estruturais, avaliados pela topografia de disco óptico e pela polarimetria de varredura a laser apresentam boa capacidade na diferenciação entre olhos normais e glaucomatosos. Entretanto, os melhores resultados são obtidos ao combiná-los em uma função linear discriminante com parâmetros derivados de ambos os instrumentos
Abstract: Purpose: To test the ability of structural parameters (as measured by the Scanning Laser Polarimetry and Confocal Scanning Laser Ophthalmoscopy) to discriminate between normal and glaucomatous eyes. Methods: After approval of the Ethics Committee of the University of Campinas, primary open-angle glaucoma patients were selected from the glaucoma service. Normal individuals were recruited from volunteers among the medical staff, university members, family and friends of patients. All individuals underwent a thorough ophthalmic evaluation, a 24-2 full threshold Humphrey visual field, Scanning Laser Polarimetry with the GDx and Confocal Scanning Laser Ophthalmoscopy with the TOPSS. Cut-off points were selected and ROC curves were created for each individual Confocal Scanning Laser Ophthalmoscopy and Scanning Laser Polarimetry parameters. Finally, multivariate discriminant formulas were developed in order to achieve a better sensitivity (Se)/specificity (Sp) ratio for the diagnosis of glaucoma, initially separately for each device, and then combining parameters from Confocal Scanning Laser Ophthalmoscopy and Scanning Laser Polarimetry. Results: One-hundred and twelve patients with primary open-angle glaucoma and 88 normal individuals were enrolled in the study. The best Scanning Laser Polarimetry parameters in differentiating normal individuals from glaucoma patients were: The Number (Se: 79,5%, Es: 81,8%, area under the ROC curve- aROC: 0,870), Maximum Modulation (Se: 83,0%, Es: 76,1%, aROC: 0,842) and Ellipse Modulation (Se: 65,2%, Es: 88,6%, aROC: 0,831), while the best Confocal Scanning Laser Ophthalmoscopy individual parameters were: Average Disc Diameter (Se: 64%, Sp: 89%, aROC: 0,824), Total Disc Area (Se: 85%, Sp: 66%, aROC: 0,802) and Cup Area (Se: 69%, Sp: 85%, aROC: 0,797). However, the multivariate analysis resulted in better results for the Scanning Laser Polarimetry (Se: 87,5%, Es: 86,4%, aROC: 0,900), Confocal Scanning Laser Ophthalmoscopy (Se: 90%, Sp: 81%, aROC: 0,910), and for the combination of parameters from both devices (Se: 93%, Es: 91%, aROC: 0,970). Conclusion: Structural parameters, as measured by the Confocal Scanning Laser Ophthalmoscopy and Scanning Laser Polarimetry have good ability to differentiate normal individuals from glaucoma patients. However, better results are obtained combining parameters from both devices in a multivariate discriminant formula
Doutorado
Oftalmologia
Doutor em Ciências Médicas

Au cours de la dernière décennie, la pathologie numérique a été améliorée grâce aux avancées des algorithmes d'analyse d'images et de la puissance de calcul. Néanmoins, le diagnostic par un expert à partir d'images histopathologiques reste le gold standard pour un nombre considérable de maladies notamment le cancer. Ce type d'images préserve la structure des tissus aussi proches que possible de leur état vivant. Ainsi, cela permet de quantifier les objets biologiques et de décrire leur organisation spatiale afin de fournir une description plus précise des tissus malades. L'analyse automatique des images histopathologiques peut avoir trois objectifs: le diagnostic assisté par ordinateur, l'évaluation de la sévérité des maladies et enfin l'étude et l'interprétation des mécanismes sous-jacents des maladies et leurs impacts sur les objets biologiques. L'objectif principal de cette thèse est en premier lieu de comprendre et relever les défis associés à l'analyse automatisée des images histologiques. Ensuite, ces travaux visent à décrire les populations d'objets biologiques présents dans les images et leurs relations et interactions à l'aide des statistiques spatiales et également à évaluer la significativité de leurs différences en fonction de la maladie par des tests statistiques. Après une étape de séparation des populations d'objets biologiques basée sur la couleur des marqueurs, une extraction automatique de leurs emplacements est effectuée en fonction de leur type, qui peut être ponctuel ou surfacique. Les statistiques spatiales, basées sur la distance pour les données ponctuelles, sont étudiées et une fonction originale afin de mesurer les interactions entre deux types de données est proposée. Puisqu'il a été montré dans la littérature que la texture d'un tissu est altérée par la présence d'une maladie, les méthodes fondées sur les motifs binaires locaux sont discutées et une approche basée sur une modification de la résolution de l'image afin d'améliorer leur description est introduite. Enfin, les statistiques descriptives et déductives sont appliquées afin d'interpréter les caractéristiques extraites et d'étudier leur pouvoir discriminant dans le cadre de l'étude des modèles animaux de cancer colorectal. Ce travail préconise la mesure des associations entre différents types d'objets biologiques pour mieux comprendre et comparer les mécanismes sous-jacents des maladies et leurs impacts sur la structure des tissus. En outre, nos expériences confirment que l'information de texture joue un rôle important dans la différenciation des deux modèles d'implantation d'une même maladie
During the last decade, digital pathology has been improved thanks to the advance of image analysis algorithms and calculus power. However, the diagnosis from histopathology images by an expert remains the gold standard in a considerable number of diseases especially cancer. This type of images preserves the tissue structures as close as possible to their living state. Thus, it allows to quantify the biological objects and to describe their spatial organization in order to provide a more specific characterization of diseased tissues. The automated analysis of histopathological images can have three objectives: computer-aided diagnosis, disease grading, and the study and interpretation of the underlying disease mechanisms and their impact on biological objects. The main goal of this dissertation is first to understand and address the challenges associated with the automated analysis of histology images. Then it aims at describing the populations of biological objects present in histology images and their relationships using spatial statistics and also at assessing the significance of their differences according to the disease through statistical tests. After a color-based separation of the biological object populations, an automated extraction of their locations is performed according to their types, which can be point or areal data. Distance-based spatial statistics for point data are reviewed and an original function to measure the interactions between point and areal data is proposed. Since it has been shown that the tissue texture is altered by the presence of a disease, local binary patterns methods are discussed and an approach based on a modification of the image resolution to enhance their description is introduced. Finally, descriptive and inferential statistics are applied in order to interpret the extracted features and to study their discriminative power in the application context of animal models of colorectal cancer. This work advocates the measure of associations between different types of biological objects to better understand and compare the underlying mechanisms of diseases and their impact on the tissue structure. Besides, our experiments confirm that the texture information plays an important part in the differentiation of two implemented models of the same disease

Objetivos: Analisar o potencial da oftalmoscopia confocal de varredura a laser, através do Heidelberg Retina Tomograph (HRT), para predizer o risco de progressão em pacientes com suspeita de glaucoma. Comparar os resultados obtidos com o índice de probabilidade de glaucoma (GPS) do HRT aos resultados da análise de regressão de Moorfields (ARM), dos parâmetros morfométricos e da avaliação das fotografias estereoscópicas. Métodos: Uma coorte foi selecionada com 223 pacientes com suspeita de glaucoma, que foram seguidos por um período médio de 64,9 ± 37,3 meses. A suspeita de glaucoma baseou-se na aparência do disco óptico e/ou na pressão intraocular elevada (> 21 mmHg). Todos os participantes apresentavam dois exames de campo visual normais ao entrar no estudo. Conceituou-se progressão como o desenvolvimento de um defeito confirmado de campo visual ou deterioração do disco óptico na avaliação seriada das estereofotografias. A associação entre os resultados do HRT na época do início do acompanhamento e a progressão para glaucoma foi investigada através de modelos de regressão do tipo Cox. Usou-se o C-index para a comparação entre os modelos com os diversos parâmetros do HRT, isolados ou ajustados para os outros já conhecidos fatores de risco para progressão (idade, espessura corneana, pressão intraocular e pattern standard deviation PSD). Resultados: No período do estudo, 46 pacientes (21%) apresentaram progressão. Na análise multivariada, o GPS, a ARM e diversos parâmetros morfométricos foram preditivos para progressão, assim como a avaliação subjetiva das estereofotografias. Cada GPS 0,1 maior foi associado com um aumento de 23% no risco de progressão (C-index de 0,69). Os resultados anormais nas classificações finais do GPS e da ARM foram associados a aumentos de 3 e 2 vezes no risco de progressão, respectivamente (C-indexes de 0,70 e 0,68, respectivamente). O parâmetro com o melhor C-index foi a área seccional tranversa da camada de fibras nervosas (0,72). Uma área 0,3 mm2 menor foi associada a um risco 62% maior de progressão. A comparação do valor preditivo entre os modelos com o GPS e com a avaliação subjetiva das estereofotografias foi similar (C-indexes de 0,69 e 0,68, respectivamente). Conclusão: Nesse estudo observamos que as análises objetivas do disco óptico e da região peripapilar obtidas com o HRT contribuíram na avaliação do risco de progressão em pacientes com suspeita de glaucoma. O GPS mostrou-se tão eficaz quanto os parâmetros morfométricos e a análise de regressão de Moorfields, e a comparação do desempenho dos modelos contendo a avaliação subjetiva das estereofotografias e aqueles contendo a avaliação objetiva pelo GPS não mostrou diferenças significativas
Purpose: To evaluate the ability of baseline confocal scanning laser ophthalmoscopy, using the Heidelberg Retina Tomograph (HRT), in predicting the development of progression in patients suspected of having glaucoma. In addition, the study also aimed to compare the predictive abilities of the glaucoma probability score (GPS) with those of the Moorfields regression analysis (MRA) and stereometric parameters, and to compare the performance of the HRT with that of subjective evaluation of optic disc stereophotographs. Methods: This longitudinal study included a cohort of 223 eyes suspected of having glaucoma, which were followed for an average of 64.9 ± 37.3 months. Included suspects had a suspicious appearance of the optic disc and/or elevated intraocular pressure, but normal visual fields. Progression was defined as the development of either repeatable abnormal visual fields or glaucomatous structural deterioration in the appearance of the optic disc during the study period. The association between baseline HRT parameters and progression was investigated by Cox regression models. The comparison between models with HRT parameters, individually or combined with other known risk factors (age, central corneal thickness, pattern standard deviation and intraocular pressure), performed by comparing their C-indexes. Results: Forty-six (21%) eyes converted during the study period. In multivariable models, the GPS, the MRA, and the stereometric parameters were all predictive of progression. A GPS 0.1 larger was associated with an increase of 23% in the risk of progression (C-index of 0.69). Abnormal final classifications for the GPS or the MRA were associated with a three-fold and two-fold increase in the risk of progression, respectively (with C-indexes of 0.70 and 0.68, respectively). The parameter with the best C-index was the nerve fiber layer cross-sectional area (0.72). An area 0.3 mm2 smaller was associated with a 62% higher risk of an individual progress. The comparison between models with the HRT parameters and those with the subjective stereophotograph evaluation had similar results (C-indexes of 0.69 and 0.68, respectively). Conclusion: In this study, we were able to show that the objective structural assessment of the optic disc and peripapillary area obtained with the HRT was significantly predictive for progression in suspected individuals. The GPS was as predictive as the other HRT parameters, and no significant differences were observed between models with the GPS and those with the subjective assessment of the stereophotographs.

Abstract Glaucoma is a group of eye diseases characterized by a chronic, progressive optic neuropathy. During the disease process, the axon damage of the retinal ganglion cells leads to changes in the retinal nerve fiber layer, causing optic nerve head, and visual field defects typical of glaucoma. The Heidelberg Retina Tomograph (HRT) is a confocal scanning laser imaging device acquiring and analysing three-dimensional data of the ocular fundus wit good accuracy and reproducibility. Conventional planimetric measurements were compared to those taken with the HRT in a pilot study of 12 eyes with early glaucomatous optic disc, retinal nerve fiber layer and/or visual field abnormalities. The neuroretinal rim area measurements and cup-to-disc area ratio did not differ statistically from each other when using these two different methods. The effect of four different reference levels on the HRT parameter measurement values was tested in two separate studies. In the first study there were 67 eyes, 40 of the eyes were healthy and 27 eyes had glaucoma of different stages. Then, 279 eyes, 180 of which were non-glaucomatous and 99 glaucoma eyes, were included in another study. The flexible reference level gave the most reliable HRT parameter measurement values in both non-glaucomatous and glaucomatous eyes. The ability of the HRT parameters to separate between non-glaucomatous and glaucomatous eyes was tested in 77 eyes, 40 of the eyes were non-glaucomatous, 10 ocular hypertensives and 27 eyes had different stages of glaucoma. The reference level dependent HRT parameters cup-to-disc area ratio, vertical linear cup-to-disc ratio, mean retinal nerve fiber layer thickness (RNFLt) and rim volume as well as the reference level non-dependent HRT parameter, cup shape measure (CSM), separated best between the clinical groups. The best combination of the HRT and other structural and functional parameters in separating between non-glaucomatous and glaucomatous eyes was studied in 55 eyes. There were 32 non-glaucomatous eyes and 23 eyes with ocular hypertension or glaucoma. CSM, RNFLt, together with age- and lens coloration-corrected mean deviation of the B/Y perimetry showed good discrimination (ROC area 0.91) between non-glaucomatous and glaucomatous eyes
Tiivistelmä Glaukooma koostuu joukosta hitaasti eteneviä näköhermon rappeumasairauksia. Sairausprosessin aikana verkkokalvon gangliosolujen aksonivaurio johtaa muutoksiin verkkokalvon hermosäiekerroksessa ja näköhermon päässä aiheuttaen glaukoomalle tyypillisiä näkökenttämuutoksia. The Heidelberg Retina Tomograph (HRT) on konfokaali laserskanneritekniikkaan perustuva kuvantamislaite, joka tuottaa ja analysoi silmänpohjasta saatua kolmiulotteista mittaustietoa tarkasti ja toistettavasti. Tavanomaisen planimetrian antamia mittaustuloksia verrattiin HRT:n antamiin tuloksiin 12:ssa silmässä, joissa oli todettu varhaisia glaukoomamuutoksia. Näköhermon pään hermoreunan (rim) pinta-ala ja keskuskuopan suhde papillan läpimittaan eivät poikenneet tilastollisesti toisistaan näitä kahta menetelmää käytettäessä. Neljän eri referenssitason vaikutusta HRT-parametrien mittausarvoihin testattiin kahdessa eri tutkimuksessa. Ensimmäisen tutkimusaineisto koostui yhteensä 67:stä silmästä, joista 40 oli terveitä ja 27:ssä eriasteisia glaukoomamuutoksia. Toisessa tutkimuksessa oli yhteensä 279 silmää, joista 180 oli terveitä ja 99:llä oli glaukooma. Papillomakulaarisäikeisiin tukeutuva, fleksiibeli referenssitaso antoi luotettavimmat HRT-parametrien mittaustulokset sekä terveissä että glaukoomasilmissä. HRT-parametrien kykyä erottaa terveet silmät glaukomatoottisista testattiin yhteensä 77:ssä silmässä, joista 40 oli terveitä, 10 oli korkeapaineisia ilman glaukoomamuutoksia, ja 27:ssä oli glaukoomamuutoksia. Referenssitasosta riippuvaiset HRT-parametrit, keskuskuopan suhde papillan läpimittaan, vertikaali-lineaarinen keskuskuopan suhde papillan läpimittaan, keskimääräinen verkkokalvon hermosäiekerroksen paksuus (RNFLt) ja `rim´:in tilavuus samoin kuin referenssitasosta riippumaton keskuskuopan ´vinous´-mitta (CSM) erottelivat parhaiten nämä kliiniset ryhmät toisistaan. Terveitä ja glaukoomasilmiä erottelevaa HRT:n ja muiden rakenteellisten ja toiminnallisten parametrien kombinaatiota etsittiin 55:n silmän aineistosta. Silmistä 32 oli terveitä ja 23 korkeapaineisia ja/tai glaukoomavaurioisia. CSM ja RNFLt, yhdessä iän ja mykiövärjäytymisen suhteen korjatun sinikeltaperimetrian keskipoikkeaman kanssa osoittivat hyvää erottelukykyä (ROC area 0.91) terveiden ja glaukoomasilmien välillä

O objetivo desse estudo foi comparar a habilidade da avaliação subjetiva do disco óptico e da camada de fibras nervosas (CFN) por oftalmologistas generalistas e por um glaucomatólogo com medidas objetivas pela tomografia de coerência óptica (Stratus OCT), oftalmoscopia confocal de varredura a laser (HRTIII), e a polarimetria de varredura a laser (GDxECC) para discriminar olhos glaucomatosos de olhos normais. 61 olhos glaucomatosos e 57 olhos normais de 118 indivíduos foram incluídos nesse estudo. Três oftalmologistas generalistas independentes e um glaucomatólogo avaliaram fotografias estereoscópicas do disco óptico. Curvas ROC (Receiver Operator Characteristic) foram construídas para cada técnica de imagem e a sensibilidade a uma especificidade fixa foi determinada. Comparações das áreas sob essas curvas (AROC) e a concordância (k) foram determinadas entre a graduação das fotografias estereoscópicas e o melhor parâmetro de cada exame de imagem computadorizado. O melhor parâmetro de cada técnica de imagem computadorizada (CFN temporal inferior do Stratus OCT = 0,92; área integrada vertical da topografia do disco óptico pelo Stratus OCT = 0,86; espessura macular do setor inferior externo fornecido pelo Stratus OCT = 0,82; NFI do GDxECC = 0,91; razão área E/D do HRT3 = 0,83) mostrou AROC maior do que a graduação das fotografias estereoscópicas por oftalmologistas generalistas (0,80) para separar olhos glaucomatosos de olhos normais. A graduação por glaucomatólogo forneceu AROC igual ou maior (0,92) do que o melhor parâmetro de exame computadorizado de imagem. A avaliação das fotografias estereoscópicas por glaucomatólogo mostrou melhor concordância com o melhor parâmetro de cada técnica de imagem quantitativa na classificação de olhos como glaucomatosos ou normais comparadas à avaliação de fotografias estereoscópicas por oftalmologistas generalistas. A combinação da avaliação subjetiva do disco óptico por oftalmologistas generalistas com parâmetros objetivos da CFN melhorou a identificação de olhos com glaucoma em uma proporção maior do que a combinação desses parâmetros objetivos com a avaliação do disco óptico por um glaucomatólogo (29,5% versus 19,7%, respectivamente). Os resultados sugerem que a habilidade diagnóstica de todos os exames de imagem computadorizados estudados mostrou melhor desempenho que a avaliação subjetiva do disco óptico por oftalmologistas generalistas, mas não por um glaucomatólogo. Medidas objetivas da CFN podem permitir um aperfeiçoamento na detecção de glaucoma quando combinados com a avaliação subjetiva do disco óptico por oftalmologistas generalistas ou por um glaucomatólogo.
The purpose of this study was to compare the ability of subjective assessment of optic nerve head and retinal nerve fiber layer by general ophthalmologists and by a glaucoma expert with objective measurements by optical coherence tomography (Stratus OCT, Carl Zeiss Meditec, Inc.), confocal scanning laser ophthalmoscope (HRT III; Heidelberg Engineering, Heidelberg, Germany), and scanning laser polarimetry (GDx ECC; Carl Zeiss Meditec, Inc., Dublin, CA) in discriminating glaucomatous and normal eyes. 61 glaucomatous and 57 normal eyes of 118 subjects were included in the study. Three independent general ophthalmologists and one glaucoma expert evaluated ONH stereophotographs. Receiver operating characteristic curves were constructed for each imaging technique and sensitivity at fixed specificity was estimated. Comparisons of areas under these curves (aROC) and agreement (k) were determined between stereophoto grading and best parameter from each technique. Best parameter from each technique showed larger aROC (StratusOCT RNFL = 0.92; StratusOCT ONH vertical integrated area = 0.86; StratusOCT macular thickness = 0.82; GDxECC = 0.91; HRT3 global cup-to-disc area ratio = 0.83) compared to stereophotograph grading by general ophthalmologists (0.80) in separating glaucomatous and normal eyes. Glaucoma expert stereophoto grading provided equal or larger aROC (0.92) than best parameter of each computerized imaging device. Stereophoto evaluated by a glaucoma expert showed better agreement with best parameter of each quantitative imaging technique in classifying eyes either as glaucomatous or normal compared to stereophoto grading by general ophthalmologists. The combination of subjective assessment of the optic disc by general ophthalmologists with RNFL objective parameters improved identification of glaucoma patients in a larger proportion than the combination of these objective parameters with subjective assessment of the optic disc by a glaucoma expert (29.5% versus 19.7%, respectively). The results suggests that the diagnostic ability of all imaging techniques showed better performance than subjective assessment of the ONH by general ophthalmologists, but not by a glaucoma expert. Objective RNFL measurements may provide improvement in glaucoma detection when combined with subjective assessment of the optic disc by general ophthalmologists or by a glaucoma expert.

Alterações no aspecto do disco óptico e da camada de fibras nervosas da retina (CFN) freqüentemente precedem o aparecimento de defeitos de campo visual no glaucoma, o que faz com que a avaliação destas estruturas seja essencial para o diagnóstico precoce e prevenção da perda visual nesta doença. A polarimetria de varredura a laser (GDx VCC), a oftalmoscopia confocal de varredura a laser (HRT II [Heidelberg Retina Tomograph]) e a tomografia de coerência óptica (Stratus OCT) são tecnologias que permitem a avaliação objetiva e quantitativa do disco óptico e da CFN. No presente estudo, estas tecnologias foram comparadas em sua habilidade para diferenciar pacientes glaucomatosos de indivíduos normais. Pacientes com glaucoma foram selecionados com base na presença de defeitos reprodutíveis de campo visual na perimetria acromática automatizada (glaucoma perimétrico), ou com base na evidência documentada de progressão do dano glaucomatoso ao disco óptico, sem presença de defeitos de campo visual (glaucoma pré-perimétrico). Indivíduos normais apresentaram campos visuais e exame clínico dentro da normalidade. Todos os indivíduos foram submetidos a exames com o GDx VCC, HRT II, Stratus OCT e campo visual dentro de um período de três meses. Diversas medidas foram utilizadas para avaliação da acurácia diagnóstica, incluindo áreas sob as curvas receiver operating characteristic (AROC), sensibilidades para especificidades fixas, e razões de probabilidade. Modelos estatísticos foram utilizados para avaliação da influência da severidade do glaucoma e tamanho do disco óptico na performance diagnóstica dos diferentes instrumentos. Um olho de cada indivíduo foi utilizado para análise. Dos 258 sujeitos inicialmente avaliados, 33 (13%) foram posteriormente excluídos por apresentarem imagens de baixa qualidade em pelo menos um dos aparelhos, restando 225 indivíduos (133 glaucomatosos e 92 normais) para análise. Na comparação entre os parâmetros de cada instrumento com maiores valores de AROC, o parâmetro do GDx VCC, Nerve Fiber Indicator (NFI; AROC = 0,91), e o parâmetro do Stratus OCT, Espessura Média (AROC = 0,90), apresentaram áreas sob as curvas ROC significativamente superiores à do parâmetro do HRT II, função discriminante de Bathija (AROC = 0,84). A severidade do defeito de campo visual exerceu influência significativa sob a acurácia diagnóstica dos três instrumentos, com melhora no poder diagnóstico em casos mais avançados da doença. Para o GDx VCC e Stratus OCT, o aumento no tamanho do disco óptico foi associado à diminuição na sensibilidade para detecção do glaucoma; enquanto que, para o HRT II, diminuição no tamanho do disco óptico foi associada à diminuição na sensibilidade. Razões de probabilidade para resultados anormais nas xxv classificações finais de cada instrumento foram associadas a grandes efeitos de mudança na probabilidade pós-teste em relação à probabilidade préteste, sugerindo que o encontro de um resultado anormal em qualquer um destes testes, durante a avaliação de um paciente com suspeita de glaucoma, tem impacto significativo em aumentar a probabilidade de que a doença esteja presente. Além disso, os resultados obtidos na avaliação de pacientes com glaucoma pré-perimétrico sugerem que todos os três instrumentos sejam capazes de detectar alterações estruturais precoces no glaucoma, antes do aparecimento de defeitos de campo visual na perimetria acromática
Changes in the structural appearance of the optic nerve head (ONH) and retinal nerve fiber layer (RNFL) have been reported to precede the development of visual field loss in glaucoma. Detection of ONH and RNFL damage is therefore crucial for early diagnosis of glaucoma and prevention of functional loss from the disease. Scanning laser polarimetry (GDx VCC), confocal scanning laser ophthalmoscopy (HRT II [Heidelberg Retina Tomograph]) and optical coherence tomography (Stratus OCT) are different technologies capable of providing objective and quantitative information related to these structures. The purpose of the present study was to compare, in a single population, the diagnostic abilities of these technologies in the discrimination of glaucomatous patients from healthy subjects. Glaucoma patients were selected based on the presence of repeatable visual field defects, as identified by standard automated perimetry (perimetric glaucoma), or documented evidence of progressive damage to the optic disc, in the absence of detectable visual field loss (preperimetric glaucoma). Normal subjects had normal visual fields and normal clinical examination. All subjects underwent imaging with the GDx VCC, HRT II and Stratus OCT within a 3-month period. Several measures were used for evaluation of diagnostic accuracy, including the area under the receiver operating characteristic curve (AROC), sensitivities at fixed specifities, and likelihood ratios. Statistical models were used to evaluate the influence of glaucoma severity and optic disc size on the diagnostic performance of the different instruments. One eye of each individual was randomly selected for statistical analysis. From an initial group of 258 eligible subjects, 33 (13%) had images of unacceptable quality, leaving 133 glaucoma patients and 92 healthy subjects for further analysis. In the comparison of the parameters with highest values of AROC from each instrument, the GDx VCC Nerve Fiber Indicator (AROC = 0.91) and the Stratus OCT Average Thickness (AROC = 0.90) perfomed significanlty better than the HRT II Bathija discriminant function (AROC = 0.84). For all instruments, the diagnostic accuracy increased with increasing severity of visual field defects. For the GDx VCC and Stratus OCT parameters, an increase in the size of the optic disc was related to a decrease in the sensitivity for glaucoma detection. An opposite effect was observed with the HRT II: a decrease in the size of the optic disc was related to a decrease in the sensitivity for glaucoma diagnosis. Abnormal results for each of the instruments were associated with strong positive likelihood ratios, indicating a large change from prestest to posttest probability of glaucoma. These results suggest that the finding of an abnormal result in any of these tests, when assessing a patient suspect of having glaucoma, would substantially raise the probability of disease. Results of the evaluation of patients with preperimetric glaucoma also suggest that all three instruments are able to detect early glaucomatous structural damage in the absence of visual field loss

There is considerable dissatisfaction with the reliability and sensitivity of the methods used to assess the glaucomatous visual field. Two types of visual field test, which have been proposed as having potential in diagnosing glaucomatous visual field defects, have been modified and tested on a group of patients from a glaucoma clinic, a group of age-matched control subjects and a younger control group. 1. A grating pattern was generated using a laser interferometer which projected a large diameter image onto the retina independent of the subject's refractive error. The experimental set up which produced the most reliable and consistently low contrast threshold values in normal subjects was sought. The display characteristics which were examined included different orientations for the field quadrants as projected to the subject; stationary and flickering patterns using a variety of flicker generation methods; red and green light sources; and concentric or vertical sinusoidal grating patterns. Ultimately the optimal display was found to be a stationary image consisting of a green, vertical sinusoidal grating pattern. Arcuate regions of the visual field (at 10 to 20° from fixation) were stimulated in 4 distinct, obliquely oriented quadrants and a low spatial frequency (one cycle per degree) was chosen. 2. Normal limits were obtained from age-matched control subjects for comparison with the results for the patients from the glaucoma clinic. In the patient group, of the 13 who completed the test, 9 individuals were identified as abnormal with one or more of their contrast threshold scores exceeding that limit. The patients' Friedmann visual field plots were analysed and the amount of loss in each quadrant was quantified. There was a positive correlation between the quantified visual field loss and contrast threshold scores in 6 patients, a statistically borderline correlation in 2 patients and the absence of a correlation was found in 5 patient's results. The results for a subgroup of 6 visually abnormal eyes (not affected by glaucoma) excluded from the age-matched control group are also described. Their visual defects included mild cataract, amblyopia and retinal detachment. There were no clear abnormal results in 5 of the eyes in this group; however, in one subject with retinal scarring due to an infection, there was a distinct elevation of contrast threshold in the affected eye. Humphrey visual field plots were obtained for all but one of the age-matched control subjects. 3. Simultaneous brightness ratios (SBR) have previously been shown to provide an indication of'glaucomatous damage. The same subject groups as described above were tested. SBRs were obtained for central vision in both eyes of subjects (inter-ocular ratio). This technique was now extended for the first time to paired regions within each eye (intra-ocular ratios) producing 'nasal I temporal' and 'upper I lower' ratios. In each test the subject controlled the brightness ratio which was changed in a smoothly graduated and continuou8 way. The most effective procedure for recording repeatable SBRs was first explored, and it was determined that these could be best obtained by alternating the start point of the graduated filter position. For each subject, 5 ratios were obtained: inter-ocular SBR; upper I lower intra-ocular SBR for right eye and left eye; and nasal I temporal intra-ocular SBR for right eye and left eye. 4. Normal limits were obtained from age-matched control subjects for comparison with the results for the patients from the glaucoma clinic. In each of the 5 SBR tests carried out, these limits were wide, reflecting considerable variation in the normal results. Of the 14 patients who completed the tests, 5 were identified as abnormal by one or more of their SBRs being outside normal limits. Three of these were identified as abnormal by their inter-ocular SBRs alone, one was abnormal according to his upper I lower intra-ocular SBR alone and one patient had an abnormal inter-ocular SBR and an abnormal intra-ocular SBR. The corresponding regions of the patients' Friedmann visual fields were quantified, and these values were used to calculate visual field loss ratios. There was a positive correlation between the visual field loss ratios and SBRs in 3 patients, but no correlation in 11 patients. In the sub-group of 6 visually abnormal eyes without glaucoma, mild cataract appeared not to adversely affect SBR. Mean SBRs were normal in the subject with retinal detachment but there was evidence of an enhanced amount of variation in the readings. Two subjects with a damaged retina and one with an amblyopic eye did produce abnormal inter-ocular SBRs, with the normal eye being significantly more sensitive in both cases.

Purpose: Optical coherence tomography (OCT) retinal nerve fiber layer (RNFL) thickness measurements are a reproducible and quantitative diagnostic modality widely used for glaucoma evaluations, but a high rate of testing artifacts limits its clinical utility. In Part I of this thesis, we aimed to characterize artifact types and assess artifact rates in two-dimensional (2D) RNFL thickness measurements obtained by the Spectralis OCT machine (Heidelberg Engineering, Heidelberg, Germany), as well as to determine patient factors and eye conditions associated with a higher artifact prevalence. In Part II of this thesis, we aimed to compare a new parameter, peripapillary retinal volume (RV), with the traditional 2D RNFL thickness parameter for diagnostic capability and artifact rates. Methods: Part I: The prevalence of 12 artifact types were described in this retrospective, cross sectional review of 2313 eye scans from 1188 patients who underwent a complete eye examination with Spectralis OCT scanning during the period of September 2009 to July 2013. Generalized estimating equations model was used to analyze associations between increased artifact prevalence and 10 patient characteristics, including age, sex, race, visual acuity, refractive error, astigmatism, cataract status, glaucoma staging, visual field reliability, and glaucoma diagnosis. Part II: This is a retrospective, cross-sectional review. A total of 180 subjects [113 open angle glaucoma (OAG) and 67 normal participants] had spectral domain optical coherence tomography (OCT) volume scans and RNFL thickness measurements (Spectralis, Heidelberg Engineering, Heidelberg, Germany). Peripapillary RV values were calculated using a custom-designed program with 4 different sized circumpapillary annuli (CA): CA1 had circle diameters of 2.5 and 3.5 mm; CA2, 3 and 4 mm; CA3, 3.5 and 4.5 mm; and CA4, 4 and 5 mm. Area under the receiver operating characteristic (AUROC) curves, sensitivity, and specificity were calculated for global, quadrant, and octant regions for RV (CA1 – CA4) and RNFL thickness. Pair-wise comparisons were conducted between RV and RNFL measurements. Artifacts rates were determined. Results: Part I: A total of 1070 or 46.3% of the 2313 2D eye scans had at least one artifact. De-centration error was the most common artifact (27.8%), followed by posterior vitreous detachment artifacts (14.4%). Visual acuity of less than 20/40 (p<0.0001), presence of moderate to severe cataracts (p<0.0001), advanced stage of glaucoma (p<0.0001), and a diagnosis of open angle glaucoma (p=0.0003) were associated with increased prevalence of artifacts. Part II: Of the 180 study subjects who had 3D eye scans, mean age was 62.6 ± 15.4 years and 41.7% were male. Among RV measurements, best diagnostic performances were for the smallest two annuli for inferior RV (CA1 0.964, CA2 0.955). Of the 4 annuli, the smallest CA1 had the highest diagnostic performance. Of specific regions, the inferior RV quadrant had the highest performance across CA1 to CA4. Peripapillary RV had similar diagnostic capability compared to RNFL thickness (p > 0.05). The artifact rate per B-scan for RV was 6.0%, and for 2D RNFL thickness scans was 32.2%. Conclusions: Clinicians should first assess scans for artifacts and pay attention to patient characteristics associated with a higher prevalence of artifacts before making therapeutic decisions based on RNFL thickness measurements. Meanwhile, the diagnostic capability of RV could be equal to that of RNFL thickness for diagnosing perimetric OAG, with fewer artifacts. RV may be a useful novel parameter in the evaluation of perimetric glaucoma.

Orientador: Vital Paulino Costa
Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Ciências Médicas
Made available in DSpace on 2018-08-27T18:27:45Z (GMT). No. of bitstreams: 1 Barella_KleytonArlindo_D.pdf: 7085227 bytes, checksum: 6f0489a5abce751aee2f5fb29f968473 (MD5) Previous issue date: 2015
Resumo: O presente estudo teve como objetivo investigar a acurácia dos classificadores de aprendizagem de máquina (MLCs) utilizando parâmetros da camada de fibras nervosas retiniana (CFNR) e do disco óptico (DO) obtidos com um tomógrafo de coerência óptica de domínio espectral (SD-OCT) para o diagnóstico de glaucoma. Cinquenta e sete pacientes com defeito de campo visual glaucomatoso leve a moderado e 46 indivíduos normais foram incluídos. Todos os 103 pacientes foram submetidos a exame oftalmológico completo, perimetria automatizada acromática (Humphrey Field Analyzer II®, Carl Zeiss Meditec Inc., Dublin, CA, EUA) e tomografia com um SD-OCT (Cirrus® SD-OCT; Carl Zeiss Meditec Inc., Dublin, CA, EUA). Obtiveram-se curvas ROC (receiver operating characteristic) para cada um dos 17 parâmetros da CFNR (espessura média, espessura nos quatro quadrantes (temporal, superior, nasal e inferior), espessura de cada uma das 12 horas de relógio) e para cada um dos os seis parâmetros do DO (área da rima, área do disco, volume da escavação, relação escavação/disco vertical, relação escavação/disco média e relação área escavação/disco). Posteriormente, foram testados os seguintes MLCs: Bagging (BAG), Naive-Bayes (NB), Support Vector Machine - Linear (SVML), Support Vector Machine - Gaussian (SVMG), Multilayer Perceptrons (MLP), Rede de Funções de Base Radial (RBF), Random Forest (RAN), Ensemble Selection (ENS), Árvore de Decisão (CTREE) e Adaptative Boosting (ADA). Compararam-se áreas sob as curvas ROC (aROCs) obtidas para cada parâmetro do SD-OCT e para cada um dos 10 MLCs. A idade média foi de 56,5 ± 8,9 anos para os indivíduos normais e 59,9 ± 9,0 anos para os pacientes glaucomatosos (p=0,054). Valores médios do "mean deviation" (MD) foram de -1,4 ± 1,6 dB para os indivíduos normais e -4,0 ± 2,4 dB para os pacientes glaucomatosos (p<0,001). Os parâmetros do SD-OCT com as maiores aROCs foram: relação área escavação/disco (0,846) e relação escavação/disco média (0,843). aROCs obtidas com os classificadores variaram de 0,687 (CTREE) a 0,877 (RAN). A aROC obtida com o classificador RAN testado com 13 parâmetros (0,877) não foi significativamente diferente da aROC obtida com o melhor parâmetro isolado do SD-OCT (0,846) (p=0,542). Os MLCs mostraram boa acurácia, mas não melhoraram significativamente a sensibilidade e especificidade do Cirrus® SD-OCT para o diagnóstico de glaucoma
Abstract: This study aimed to investigate the diagnostic accuracy of machine learning classifiers (MLCs) using retinal nerve fiber layer (RNFL) and optic nerve head (ONH) parameters obtained with spectral domain optical coherence tomography (SD-OCT) for the diagnosis of glaucoma. Fifty-seven patients with early to moderate glaucomatous visual field damage and 46 normal individuals were recruited. All 103 individuals underwent a complete ophthalmological examination, achromatic standard automated perimetry (Humphrey Field Analyzer II®, Carl Zeiss Meditec Inc., Dublin, CA, USA) and tomography with SD-OCT (Cirrus® SD-OCT; Carl Zeiss Meditec Inc., Dublin, CA, USA). Receiver operating characteristic (ROC) curves were built for each one of the 17 RNFL parameters (average thickness, four quadrants (temporal, superior, nasal and inferior), 12 clock-hour measurements) and each one of the six ONH parameters (rim area, disc area, cup volume, vertical cup/disc ratio, average cup/disc ratio and cup/disc area ratio). Subsequently, the following MLCs were tested: Bagging (BAG), Naive-Bayes (NB), Support Vector Machine ¿ Linear (SVML), Support Vector Machine - Gaussian (SVMG), Multilayer Perceptrons (MLP), Radial Basis Functions (RBF), Random Forest (RAN), Ensemble Selection (ENS), Classification Tree (CTREE) and Adaptative Boosting (ADA). Areas under ROC curves (aROCs) obtained for each SD-OCT parameter and 10 MLCs were compared. The mean age was 56,5 ± 8,9 years for normal individuals and 59,9 ± 9,0 years for glaucoma patients (p=0,054). Mean MD (mean deviation) values were -1,4 ± 1,6 dB for normal individuals and -4,0 ± 2,4 dB for glaucoma patients (p<0,001). SD-OCT parameters with the greatest aROCs were: cup/disc area ratio (0,846) and average cup/disc ratio (0,843). aROCs obtained with classifiers varied from 0,687 (CTREE) to 0,877 (RAN). The aROC obtained with RAN (0,877) tested with 13 parameters was not significantly different from the aROC obtained with the best single SD-OCT parameter (0,846) (p=0,542). MLCs showed good diagnostic accuracy but did not improve the sensitivity and specificity of Cirrus® SD-OCT for the diagnosis of glaucoma
Doutorado
Oftalmologia
Doutor em Ciências Médicas

Orientadores: Jose Paulo Cabral de Vasconcellos, Vital Paulino Costa
Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Ciencias Medicas
Made available in DSpace on 2018-08-12T19:09:11Z (GMT). No. of bitstreams: 1 Medina_FlavioMacCord_M.pdf: 2280525 bytes, checksum: d64f96bf19c317e1e22a704a8562d282 (MD5) Previous issue date: 2009
Resumo: Objetivo: Avaliar a reprodutibilidade do teste de sobrecarga hídrica (TSH) em diferentes horários em que é realizado, em pacientes com glaucoma primário de ângulo aberto (GPAA) e em indivíduos normais. Métodos: Quinze pacientes com GPAA e 30 indivíduos normais foram submetidos a três TSHs, realizados em diferentes horários do dia (às 7:00, 12:00 e 17:00), em três dias diferentes. Foram comparados os resultados dos testes em pacientes com GPAA e indivíduos normais. Foram analisadas a concordância e a correlação entre os valores de medida basal, pico e variação de pressão intra-ocular (PIO) (pico de PIO - PIO basal) nos testes realizados nos diferentes horários. Apenas as medidas do olho direito foram analisadas. Resultados: Os valores médios de medida basal, pico e variação de PIO foram significativamente maiores nos pacientes glaucomatosos que nos indivíduos normais, em todos os horários (p<0,05). A análise de Bland-Altman apresentou limites de concordância de pico e variação de PIO maiores do que o clinicamente aceitável (> 3 mmHg), apesar de o teste de Pearson demonstrar boa correlação entre os resultados. Conclusões: O TSH apresenta valores de pico e variação de Pio significativamente maiores em pacientes glaucomatosos que em indivíduos normais. Os baixos níveis de concordância entre os TSHs realizados em diferentes horários do dia sugerem uma baixa reprodutibilidade do TSH, que pode limitar sua aplicabilidade para diagnóstico e acompanhamento do glaucoma
Abstract: Purpose: To evaluate the reproducibility of the water drinking test (WDT) performed at different times of the day, in primary open angle glaucoma (POAG) patients and normal individuals. Methods: Fifteen patients with POAG and 30 normal individuals underwent three WDTs at different times of the day (7 AM, 12 PM, and 5 PM) on 3 different days. Test results in POAG patients and normal individuals were compared. Agreement and correlation of intraocular pressure (IOP) baseline levels, peak levels, and IOP change (peak IOP - baseline IOP) on tests performed at different times were evaluated. Only right eye measurements were analyzed. Results: Mean baseline IOP, peak IOP and IOP change were significantly higher in POAG patients than in normal individuals, at all time intervals (p<0.05). The Bland-Altman analysis demonstrated limits of agreement for IOP peak levels and IOP changes larger than the clinically acceptable (> 3 mmHg), even though Pearson's test revealed good correlation among the results. Conclusions: The mean IOP peak and mean IOP change observed during the WDT are significantly higher in POAG patients than in control individuals. Low levels of agreement among WDTs performed at different times of the day suggest a poor reproducibility of WDT, which may limit its applicability for the diagnosis and follow-up of glaucoma
Mestrado
Oftalmologia
Mestre em Ciências Médicas

INTRODUÇÃO: O glaucoma de ângulo fechado é reconhecido como uma das principais causas de cegueira mundial. A iridotomia representa o tratamento de eleição nos casos com fechamento angular primário, entretanto, este procedimento pode não ser suficiente para proporcionar abertura do ângulo irido-corneano, controle da pressão intra-ocular (PIO) e estabilização do processo da lesão glaucomatosa. O presente estudo tem como objetivo avaliar a morfologia do segmento anterior do olho em uma amostra de pacientes brasileiros, e realizar uma comparação entre olhos normais (sem sinais de fechamento angular prévio) e olhos com ângulos oclusíveis, antes e após a realização da iridotomia. MÉTODOS: Realizou-se um estudo prospectivo observacional tipo caso-controle em pacientes da Clínica Oftalmológica do HC - FMUSP, onde 40 olhos com ângulos oclusíveis (grupo caso) e 27 olhos normais (grupo controle) foram examinados durante o período de agosto de 2003 a dezembro de 2004. Os pacientes do grupo controle foram pareados por idade, sexo e raça. Os exames de gonioscopia e biometria ultra-sônica foram utilizados para comparar os 27 olhos normais com os 40 olhos com ângulos oclusíveis, antes e após a realização da iridotomia. A biometria ultra-sônica mediu o comprimento axial do olho (CAX), a profundidade da câmara anterior (PCA) e o diâmetro ântero-posterior do cristalino (DAPC). Os 27 olhos do grupo controle foram comparados, através da biomicroscopia ultra-sônica, antes e após a iridotomia com os 31 olhos do grupo caso, que não apresentavam goniosinéquias no quadrante inferior. As imagens da UBM foram obtidas em cortes radiais sobre típicos processos ciliares, no claro e no escuro. A distância da abertura angular a 500?m do esporão escleral (DAA500), profundidade da câmara anterior (PCA-UBM), distância do trabeculado aos processos ciliares (DTPC), espessura da íris a 500?m do esporão escleral (EI500) e distância do esporão escleral à inserção da íris (linha X) foram medidas nas imagens da UBM obtidas às 6 horas. As freqüências de processos ciliares longos sem sulco ciliar e fechamento angular aposicional no escuro também foram determinadas nessas imagens. RESULTADOS: Os parâmetros morfológicos dos 27 olhos do grupo controle apresentaram diferenças significativas quando comparados com os 40 olhos do grupo caso. Os olhos normais apresentaram ângulo irido-corneano mais aberto, menor DAPC e maiores CAX e PCA. Nas imagens da UBM os 27 olhos normais apresentaram maior DAA500, PCA-UBM, linha X, e também, maior DTPC que os 31 olhos com ângulos oclusíveis (651 ± 119 ?m e 508 ± 116 ?m; p < 0.001); porém, a EI500 não apresentou diferença significativa entre os dois grupos. Após a realização da iridotomia foi observado uma abertura significativa do seio camerular, e uma diminuição da freqüência de fechamento angular aposicional nas imagens da UBM obtidas no escuro (28/31 para 16/31). Processos ciliares longos sem sulco ciliar foram observados em 61% (19/31) dos olhos do grupo caso após a iridotomia e em 33% (9/27) dos olhos do grupo controle. CONCLUSÃO: A presença de processos ciliares longos sem sulco ciliar foi um achado comum não somente nos olhos com ângulos oclusíveis como também nos olhos normais. No entanto, nos olhos do grupo caso, os processos ciliares estavam localizados, em média, numa posição mais anterior. Após a iridotomia, mais da metade dos olhos com ângulos oclusíveis continuaram apresentando fechamento angular aposicional na UBM. Os valores preditivos da presença de fechamento angular aposicional (associada ou não a processos ciliares longos sem sulco ciliar) na detecção de pacientes sob risco de apresentarem episódios de fechamento angular precisam ser avaliados em estudos futuros
Introduction: Angle closure glaucoma is emerging as one of the leading cause of worldwide blindness. Laser iridotomy (LI) has been proposed as first line therapy for patients with angle closure, however this procedure may not be effective opening the irido-corneal angle, controlling intra-ocular pressure (IOP) and halting glaucoma progression in all cases. Our study aimed to evaluate anterior segment morphology on a cohort of Brazilian patients comparing normal eyes (no signs of angle closure) to angle closure eyes before and after LI. Methods: In this prospective observational case control study, performed from August of 2003 to December of 2004, we evaluated 40 angle closure eyes and 27 normal control eyes with no signs of angle closure at clinical exam, paired by age, race and gender. We used gonioscopy and A-scan biometry to compare anterior segment morphology of 27 normal control eyes to 40 angle closure eyes of patients from our service, before and after LI. We also used ultrasound biomicroscopy (UBM) exam, to compare 27 normal control eyes to 31 of 40 angle closure eyes with no goniosynachiae at the inferior quadrant evaluated by gonioscopy, before and after LI. Immersion 50-MHz high-frequency ultrasound transducer was used to obtain UBM images in radial scans through a typical ciliary process, in both standard light and dark conditions. A-scan biometry measured axial length (AXL), anterior chamber depth (ACD) and lens thickness (LENS). The angle opening distance at 500?m from the scleral spur (AOD500), trabecular ciliary process distance (TCPD), iris thickness at 500?m and the distance from scleral spur to iris insertion (line X) were measured at UBM images obtained at the inferior quadrant. The frequency of appositional angle closure and the presence of long ciliary process with no ciliary sulcus were also evaluated at UBM images. Results: At gonioscopy and A-scan biometry exam, 27 normal eyes had a significant wider iridocorneal angle opening, a thinner LENS and a greater AXL, ACD than angle closure eyes. At UBM exam, 27 normal control eyes had an significant wider AOD500, line X and also, a longer TCPD than angle closure eyes (651 ± 119 ?m vs. 508 ± 116 ?m; p < 0.001); however no differences were observed in iris thickness between the two groups. After LI, we observed a significant irido-corneal angle opening and the number of angle closure eyes with UBM appositional angle closure in dark condition decreased from 28/31 to 16/31. A long ciliary processes with no ciliary sulcus were observed in 61% (19/31) of angle closure eyes after LI, and also in 33% (9/27) of normal control eyes. Conclusion: A long ciliary processes and absence of ciliary sulcus were a quite common finding not only in angle closure eyes, but also in normal control eyes. However, ciliary processes were located more anteriorly in angle closure eyes. On this cohort of Brazilian patients, more than half of studied eyes submitted to LI maintained UBM appositional angle closure. Whether this apposition with or without long ciliary process and absence of ciliary sulcus detected at UBM images after LI is associated to further goniosynachiae formation and/or loss of IOP control remains to be evaluated

Glaucoma é uma neuropatia óptica cuja progressão gera comprometimento no campo visual e cegueira. Devido aos danos irreversíveis, a identicação precoce e correta é vital para o controle da progressão da doença. Para diagnóstico de glaucoma , oftalmologistas analisam dados de campo visual e da anatomia ocular obtidos através de testes. Para reduzir a quantidade de resultados falso negativos e falso positivos, diversas técnicas tem sido desenvolvidas para incrementar a sensibilidade e especicidade dos testes diagnósticos de glaucoma. Para aplicações médicas, existem estudos que indicam como uma abordagem promissora o uso de classicadores de aprendizagem de máquina baseados em diferentes conceitos como Árvores de Decisão, Redes Neurais Articiais e abordagem Bayesiana. Apesar da disponibilidade de sosticados algoritmos para desenvolvimento de classicadores, o sucesso do treinamento destes classicadores é altamente dependente de dados de qualidade. Dados de qualidade signicam que os exemplos fornecidos para treinamento devem representar a maior quantidade possível de situações encontradas no mundo real. Estes requisitos são normalmente atendidos se dados de uma grande quantidade de pacientes estiver disponível. Entretanto, diversos fatores como o perl da população avaliada, a duração das tarefas de coleta de dados, disponibilidade de equipamentos e de prossionais de saúde, além do comprometimento dos pacientes com o programa de pesquisa, restringem o tamanho do conjunto de dados. Uma possível estratégia para resolver o problema da escassez de dados para a tarefa de treinamento dos classicadores é o emprego de dados articiais que representam populações reais. Estes dados articiais serão adequados ao treinamento dos classicadores se possuírem características estatísticas de populações reais. O uso de populações articiais possibilitará a criação de conjuntos de dados com número adequado de pacientes, sem gastar anos coletando dados. Neste trabalho é apresentado um gerador de dados articiais denominado GLOR, baseado em um método de Monte Carlo, que é adequado ao treinamento de classicadores para o diagnóstico de glaucoma. A população gerada é caracterizada por dados funcionais e estruturais fornecidos pelos instrumentos de perimetria computadorizada padrão ou Standard Automated Perimetry (SAP) e tomograa de coerência óptica de alta denição ou High Denition Optical Coherence Tomography (HD-OCT). Os resultados experimentais, obtidos após o treinamento de uma Rede Neural Articial empregando população gerada pelo GLOR contendo 4500 indivíduos normais e 500 glaucomatosos e avaliação com dados de uma população real formada por 44 indivíduos normais e 26 glaucomatosos, foram: acurácia total de 87,1%, sensibilidade de 80,8%, especicidade de 90,9% e área sob curva ROC de 0,941. Tais resultados indicam que o GLOR pode ser empregado satisfatoriamente no desenvolvimento de novos métodos que possam elevar a sensibilidade e especicidade no diagnóstico de glaucoma.
Glaucoma is an optical neuropathy, whose progression results in visual eld impairments and blindness. Due to its irreversible damages, early and correct identication is very important to control glaucoma\'s progression. For glaucoma diagnosis, ophthalmologists analyze patient\'s visual eld and eyes structural data obtained by using eyes\' test equipments. In order to reduce the quantity of false-negative and false-positive results, several new techniques have been developed to increment the sensitivity and specicity of glaucoma diagnostic tests. A promising approach is the use of machine learning classiers. Classiers based on dierent concepts like Decision Trees, Articial Neural Networks, and Bayesian approach, have been developed for medical applications. Despite the availability of sophisticated algorithms for classiers development, successful training of classiers is highly dependent of good training data. Good data means that examples provided for classier training should represent the many dierent situations found in real world. These requirements are usually accomplished if data from a large number of patients is available. However, several factors like prole of evaluated population, duration of the data acquisition activities, existence of healthcare professionals, and equipment availability, and people\'s commitment to the research program, restrict the size of patient\'s dataset. A possible approach to overcome the lack of patient\'s data to perform the classier\'s training task is to use articial data that represent a real population. This articial data would be suitable for classiers training if it has similar statistical properties of a real population. The use of articial population will enable the creation of datasets with required number of patients, and without spending years measuring patients. It will also be possible to simulate scenarios and strategies before a long term research program starts. In this work is presented an articial data generator named GLOR, based on a Monte Carlo method, and suitable for the training of classiers for glaucoma diagnosis. The generated population is characterized by eyes\'functional and structural data provided by Standard Automated Perimetry (SAP) and High Denition Optical Coherence Tomography (HD-OCT) instruments. The experimental results, obtained after an Articial Neural Network training employing a population generated by GLOR comprising of 4500 normal and 500 glaucomatous individuals and evaluated by using real population data from 44 normal and 26 glaucomatous subjects, were: 87.1% for overall accuracy, 80.8% for sensitivity, 90.9% for specicity and 0.941 for the area under ROC curve. These results show that GLOR can be used as a promising approach to accelerate the development of new methods to increment sensitivity and specicity of glaucoma diagnosis.

This work deals with mathematical description of nerve fiber bundle trajectories for the diagnosis of glaucoma. Also gives a brief explanation in the principle of fundus camera and glaucoma. Its aim is to implement the model into MATLAB software, to project a~methodics of its use and also to realize it.

Orientador: Vital Paulino Costa
Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Ciências Aplicadas
Made available in DSpace on 2018-08-24T05:07:57Z (GMT). No. of bitstreams: 1 VidottiPimenta_VanessaGoncalves_D.pdf: 6967367 bytes, checksum: 6f865b9b0ad4158ca43a18b2ff974544 (MD5) Previous issue date: 2014
Resumo: O presente estudo teve como objetivo investigar a sensibilidade e especificidade dos classificadores de aprendizagem de máquina (MLC) e da tomografia de coerência óptica de domínio espectral (SD-OCT) para o diagnóstico de glaucoma. Sessenta e dois pacientes com glaucoma e defeito de campo visual inicial a moderado, e 48 indivíduos saudáveis foram incluídos. Todos os participantes foram submetidos a um exame oftalmológico completo, perimetria automatizada acromática e medida da espessura da camada de fibras nervosas da retina com SD-OCT (Cirrus HD-OCT; Carl Zeiss Meditec, Inc., Dublin, Califórnia, EUA). Uma curva ROC (receiver operating curve) foi obtida para cada parâmetro do SD-OCT: espessura média, espessura nos quatro quadrantes (temporal, superior, nasal e inferior) e a espessura a cada hora de relógio. Posteriormente, foram testados os seguintes MLCs construídos a partir dos dados do OCT: Árvore de Decisão (CTREE), Random Forest (RAN), Bagging (BAG), Adaptative Boosting (ADA), Ensemble Selection (ENS), Multilayer Perceptrons (MLP), Rede de Funções de Base Radial (RBF), Naive-Bayes (NB), Support Vector Machine - Linear (SVMl) e Support Vector Machine - Gaussian (SVMg). Áreas sob as curvas ROC (aROCs) foram obtidas para cada parâmetro do SD-OCT e para cada MLC. A média de idade foi de 57,0 ± 9,2 anos para os indivíduos saudáveis e de 59,9 ± 9,0 anos para os pacientes com diagnóstico de glaucoma (p=0,103). Valores do "mean deviation¿ (MD) foram de -4,1 ± 2,4 dB para pacientes com glaucoma e -1,5 ± 1,6 dB para indivíduos normais (p<0,001). Os parâmetros do SD-OCT com as maiores aROCs foram: quadrante inferior (0,813), espessura média global (0,807), 7 horas (0,765) e 6 horas de relógio (0,754). As aROCs dos classificadores variaram de 0,785 (ADA) a 0,818 (BAG). A aROC obtida com o BAG não foi significativamente diferente da aROC obtida com o melhor parâmetro isolado do SD-OCT (p=0,932). O SD-OCT mostrou boa acurácia diagnóstica em um grupo de pacientes com glaucoma inicial a moderado. Os MLCs não melhoraram a sensibilidade e a especificidade do SD-OCT para o diagnóstico do glaucoma
Abstract: This study aimed to investigate the sensitivity and specificity of machine learning classifiers (MLC) and spectral domain optical coherence tomography (SD-OCT) for the diagnosis of glaucoma. Sixty-two patients with early to moderate glaucomatous visual field damage and 48 healthy individuals were included. All subjects underwent a complete ophthalmologic examination, achromatic standard automated perimetry, and RNFL imaging with SD-OCT (Cirrus HD-OCT; Carl Zeiss Meditec, Inc., Dublin, California, USA). Receiver operating characteristic (ROC) curves were obtained for all SD-OCT parameters: average thickness, four quadrants (temporal, superior, nasal and inferior) and 12 clock hour measurements. Subsequently, the following MLCs were tested: Classification Tree (CTREE), Random Forest (RAN), Bagging (BAG), AdaBoost M1 (ADA), Ensemble Selection (ENS), Multilayer Perceptron (MLP), Radial Basis Function (RBF), Naive-Bayes (NB), and Support Vector Machine (SVM). Areas under the ROC curves (aROCs) were obtained for each parameter and each MLC. The mean age was 57.0 ± 9.2 years for healthy individuals and 59.9 ± 9.0 years for glaucoma patients (p=0.103). Mean deviation values were ¿4.1 ± 2.4 dB for glaucoma patients and ¿1.5 ± 1.6 dB for healthy individuals (p<0.001). The SD-OCT parameters with the greater aROCs were inferior quadrant (0.813), average thickness (0.807), 7 o¿clock position (0.765), and 6 o¿clock position (0.754). The aROCs obtained with classifiers varied from 0.785 (ADA) to 0.818 (BAG). The aROC obtained with BAG was not significantly different from the aROC obtained with the best single SD-OCT parameter (p=0.932). SD-OCT showed good diagnostic accuracy in a group of patients with early to moderate glaucoma. In this series, MLCs did not improve the sensitivity and specificity of SD-OCT for the diagnosis of glaucoma
Doutorado
Oftalmologia
Doutora em Ciências Médicas

Tato práce popisuje přehled standardních a pokročilých metod používaných k diagnose glaukomu v ranném stádiu. Na základě teoretických poznatků je implementován internetově orientovaný informační systém pro oční lékaře, který má tři hlavní cíle. Prvním cílem je možnost sdílení osobních dat konkrétního pacienta bez nutnosti posílat tato data internetem. Druhým cílem je vytvořit účet pacienta založený na kompletním očním vyšetření. Posledním cílem je aplikovat algoritmus pro registraci intenzitního a barevného fundus obrazu a na jeho základě vytvořit internetově orientovanou tři-dimenzionální vizualizaci optického disku. Tato práce je součásti DAAD spolupráce mezi Ústavem Biomedicínského Inženýrství, Vysokého Učení Technického v Brně, Oční klinikou v Erlangenu a Ústavem Informačních Technologií, Friedrich-Alexander University, Erlangen-Nurnberg.

The master thesis is focused on creating a methodology for quantification of the nerve fiber layer on photographs of the retina. The introductory part of the text presents a medical motivation of the thesis and mentions several studies dealing with this issue. Furthermore, the work describes available textural features and compares their ability to quantify the thickness of the nerve fiber layer. Based on the described knowledge, the methodology to make different regression models enabling prediction of the retinal nerve fiber layer thickness was developed. Then, the methodology was tested on the available image dataset. The results showed, that the outputs of regression models achieve a high correlation between the predicted output and the retinal nerve fiber layer thickness measured by optical coherence tomography. The conclusion discusses an usability of the applied solution.

Fundus kamera je široce dostupné zobrazovací zařízení, které umožňuje relativně rychlé a nenákladné vyšetření zadního segmentu oka – sítnice. Z těchto důvodů se mnoho výzkumných pracovišť zaměřuje právě na vývoj automatických metod diagnostiky nemocí sítnice s využitím fundus fotografií. Tato dizertační práce analyzuje současný stav vědeckého poznání v oblasti diagnostiky glaukomu s využitím fundus kamery a navrhuje novou metodiku hodnocení vrstvy nervových vláken (VNV) na sítnici pomocí texturní analýzy. Spolu s touto metodikou je navržena metoda segmentace cévního řečiště sítnice, jakožto další hodnotný příspěvek k současnému stavu řešené problematiky. Segmentace cévního řečiště rovněž slouží jako nezbytný krok předcházející analýzu VNV. Vedle toho práce publikuje novou volně dostupnou databázi snímků sítnice se zlatými standardy pro účely hodnocení automatických metod segmentace cévního řečiště.

Familial amyloid polyneuropathy (FAP) is a hereditary condition characterized by systemic deposition of transthyretin (TTR), which causes debilitating peripheral polyneuropathy, cardiopathy, nephropathy and usually after a few years, ophtalmopathy. Occasionally the onset can be atypical and the diagnosis of FAP is reliant on identifying ocular amyloid deposition clinically and histopathologically. However, images of TTR derived from the eye, identified using immunolabeling techniques, have so far not been published. In ocular tissues FAP can cause sight-threatening complications such as glaucoma and retinal amyloid angiopathy. Glaucoma in FAP often requires surgical treatment. Nonpenetrating deep sclerectomy (NPDS) is a surgical technique with several advantages over the traditional trabeculectomy. It is successfully performed in primary and many types of secondary open-angle glaucoma, but so far with limited reports in FAP. Retina imaging modalities, such as optic coherence tomography (OCT) and autofluorescence (AF), have significant value in the assessment of retinal pathologies. Fluorescein angiography is the conventional method of evaluating retinal vasculature, but requires injection of fluorescein, which has several side effects and contraindications. Recently a new non-invasive modality, the OCT angiography (OCT-A) has become a useful tool for visualizing posterior pole blood circulation. In patients with FAP the use of OCT-A has so far not been reported and only few cases of AF findings were published. This doctoral thesis, presented as a compendium of publications, is divided into three parts. The first part (Paper 1) aims to present the immunostaining images of TTR amyloid derived from the vitreous of a series of patients with FAP, which demonstrates vitreous biopsy as a valid diagnostic tool, especially in clinically challenging cases. The second part (Paper 2) is a retrospective review of clinical charts of patients with FAP to determine the prevalence and characteristics of open-angle glaucoma secondary to FAP. It reveals the particularly quick progression of glaucoma in FAP and its increased risk in patients with a previous vitrectomy. Surgical management and outcomes of the affected patients are presented, indicating that NPDS is a safe and effective treatment of glaucoma secondary to FAP. The third part (Paper 3) is an observational cross-sectional study of retinal findings in patients with FAP. It gives a descriptive analysis of retinal images in FAP using novel non-invasive techniques: AF, OCT, OCT-A, and ultra-wide-field (UWF) retinography. These modalities can be used to detect perivascular retinal amyloid deposits, as well as microvascular changes including areas of non-perfusion, allowing better understanding of the pathology, complications and prognosis of patients with FAP. It also shows that amyloid retinopathy is more frequent than previously reported. The thesis outcomes emphasize glaucoma and retinopathy as the severe irreversible complications of FAP and need for addressing them promptly. This is especially important in establishing adequate regular eye reviews in patients with FAP and identifying those individuals requiring stricter ophthalmological care to prevent vision loss.
La polineuropatía amiloidótica familiar (PAF) es una enfermedad hereditaria caracterizada por el depósito sistémico de transtiretina (TTR), que resulta en polineuropatía periférica debilitante, cardiopatía, nefropatía y, habitualmente, después de unos años, oftalmopatía. Ocasionalmente, el inicio puede ser atípico y el diagnóstico de PAF depende de la identificación de depósitos de amiloide en tejidos oculares clínicamente e histopatológicamente. Sin embargo, hasta ahora no se han publicado imágenes de TTR derivadas del ojo, identificadas utilizando técnicas de inmunotinción. En los tejidos oculares, la PAF puede causar complicaciones amenazantes para la vista, como el glaucoma y la angiopatía amiloide de la retina. El glaucoma en la PAF frecuentemente requiere tratamiento quirúrgico. La esclerectomía profunda no penetrante (EPNP) es una técnica quirúrgica con varias ventajas sobre la trabeculectomía tradicional. Se realiza con éxito en glaucoma de ángulo abierto primario y muchos tipos de glaucoma secundario, pero hasta ahora con pocos casos descritos en PAF. Las modalidades de imágen de retina, como la tomografía de coherencia óptica (OCT) y la autofluorescencia (AF), tienen un valor importante en la evaluación de las patologías retinianas. La angiografía con fluoresceína es el método convencional para evaluar la vasculatura retiniana, pero requiere la inyección de fluoresceína, que tiene varios efectos secundarios y contraindicaciones. Recientemente, una nueva modalidad no invasiva, la angiografía OCT (OCT-A) se ha convertido en una herramienta útil para visualizar la circulación sanguínea del polo posterior. En pacientes con PAF, el uso de OCT-A no ha sido publicado hasta ahora, y solo se han descrito dos casos de hallazgos de AF. Esta tesis doctoral, presentada como un compendio de publicaciones, se divide en tres partes. La primera parte (Artículo 1) tiene como objetivo presentar las imágenes de inmunotinción de TTR amiloide derivado del vítreo en una serie de pacientes con PAF, lo que demuestra que la biopsia vítrea es una herramienta de diagnóstico válida, especialmente en casos clínicamente atípicos. La segunda parte (Artículo 2) es una revisión retrospectiva de las historias clínicas de pacientes con PAF para determinar la prevalencia y las características del glaucoma de ángulo abierto secundario a la PAF. Revela la progresión particularmente rápida del glaucoma en la PAF y su mayor riesgo en pacientes con vitrectomía previa. Se ha presentado el tratamiento quirúrgico y los resultados de los pacientes afectados, lo que indica que EPNP es un tratamiento seguro y efectivo para el glaucoma secundario a PAF. La tercera parte (Artículo 3) es un estudio transversal observacional de hallazgos retinianos en pacientes con PAF. Se expone un análisis descriptivo de las imágenes de la retina en PAF utilizando nuevas técnicas no invasivas: AF, OCT, OCT-A y retinografía de campo amplio (UWF). Estas modalidades se pueden utilizar para detectar depósitos amiloides perivasculares de la retina, así como cambios microvasculares que incluyen áreas de no perfusión, lo que permite una mejor comprensión de la patología, las complicaciones y el pronóstico de los pacientes con PAF. También se muestra que la retinopatía amiloidea es más frecuente de lo que se publicó anteriormente. Los resultados de la tesis enfatizan el glaucoma y la retinopatía como las complicaciones irreversibles graves de la PAF y la necesidad de abordarlos precozmente. Esto es especialmente importante para establecer revisiones oculares regulares adecuadas en pacientes con PAF e identificar a aquellas personas que requieren atención oftalmológica más estricta para prevenir la pérdida de visión.

Objetivo: Estudar as medidas de espessura da camada de fibras nervosas da retina(CFNR) peripapilar obtidas com as tomografias de coerência óptia (oCT) time domain (TD) e spectral domain (SD) em pacientes com perda assimétrica glaucomatosa de hemicampo visual, compará-las entre si e com aquelas de controles normais. Métodos: Trinta e seis pacientes com glaucoma primário de ângulo aberto e perda de campo visual em um hemicampo (afetado) e ausência de perda no hemicampo oposto (não afetado), e 36 controles pareados por idade tiveram o olho de estudo examinado com Stratus-OCT (Carl Zeiss Meditec Inc., Dublin, Califoprnia, USA) e o 3DOCT-1000 (Topcdon, Tokyo, Japan). As medidas de espessura da CFNR peripapilar e a classificação normativa fornecida pelos aparelhos foram registrados para análise. A média aritmética dos valores do mapa total deviation em cada hemicampo (mean deviation do hemicampo) foi calculada para cada indivíduo. \"Ìndices de assimetria\" para o campo visual e para a CFNR foram calculados como a razão entre o mean deviation dos hemicampos afetado e não-afetado, e como razão entre a espessura da CFNR das hemirretinas afetada e não-afetada, respectivamente. As variáveis contínuas foram comparadas usando os testes de Mann-Whitney, Kruskal-Wallis ou Wilcoxon, quando apropriados. As variáveis categóricas foram comparadas usando o teste qui-quadrado de Pearson. O coeficiente de correlação de Spearman foi usado para testar as correlações entre as medidas de espessura da CFNR fornecidas pelos OCTs. A presença de afinamento da CFNR foi estabelecida com base nos dados normativos fornecidos pelos softwares dos OCTs. As espessuras de CFNR fora do intervalo de previsão de 95% para a mesma faixa etária foram consideradas anormais. Resultados: As medidas de CFNR corespondentes a hemicampos não-afetados [média (DP) 87,0 (17,1) um e 84,3 (20,2) um, para TD e SD-OCT, respectivamente] foram menores do que as dos controles [média (DP) 119,0 (122,2)um e 117,0 (17,7) um, para TD e SD-OCT, respectivamente, P < 0,001, para ambos]. O banco de dados normativo classificou como alterado 42% e 67% das hemirretinas correspondentes a hemicampos não-acometidos com TD e SD-OCT, respectivamente (P = 0,01). As medidas da CFNR foram consistentemente mais espessas com TD comparadas com SD-OCT. Os índices de assimetria da CFNR em pacientes com glaucoma foram semelhantes entre TD [média (DP) 0,76 (0,17)] e SD-OCT [média (DP) 0,79 (0,12), P = 0,89] e significantemente maiores do que o índice de assimetria do campo visual [média (DP) 0,36 (0,20), P < 0,001]. Conclusões: Os hemicampos normais de pacientes com glaucoma apresentaram CFNR mais fina do que de olhos saudáveis. As medidas da CFNR foram mais espessas com TD do que com SD-OCT, o qual por sua vez detectou anormalidades na espessura da CFNR mais frequentemente do que o TD-OCT
Objective: To study the peripapillary retinal nerve fiber layer (RNFL) thickness measurements obtained with time domain (TD) and spectral domain (SD) optical coherence tomography (OCT) in glaucomatous patients with asymmetric visual hemifield loss, to compare themselves and with those obtained from normal controls. Methods : Thirty -six patients with primary open-angle glaucoma with visual primary open-angle glaucoma with visual field loss in one hemifield (affected ) and absence of loss in other (non-affected), and 36 age-matched healthy controls had the eye study imaged with Stratus-OCT (Carl Zeiss Meditec Inc., Dublin, Califoprnia, USA) and 3D OCT-1000 (Topcon , Tokyo, Japan). Peripapillary RNFL thickness measurements and normative classification were recorded for analysis. Total deviation values were averaged for each hemifield (hemifield mean deviation) for ecch subject. Visual field and RNFL \"asymmetry indexes\" were calculated as the ratio between the mean deviation of affected versus non-affected hemifields and RNFL thickness between as affected versus non-affected hemiretinas, respectively. Continuous variables were compared using the Mann-Whitney, Kruskal-Wallis or Wilcoxon tests, when appropriate. Categorical variables were compared using the Pearson\'s chi-square test. The Spearman\'s rank correlation coefficient was used to test correlations between RNFL thickness measurements provided by both OCTs . The presence of RNFL thinning was establised based on normative data provided by the OCT\'s software. The RNFL thicknesses outside the prediction interval of 95% for the same age group were considered abnormal. Results: The RNFL measurements in non-affected hemifields [mean (SD 87.0 (17.1) e 84.3(20.2) um, for TD and SD-OCT, respectively] were thinner than those of normal controls [mean (SD) 119.0 (12.2) um and 117.0 (17.7) um, for TD and SD-OCT, respectively, P < 0.001 for both ] . The OCT normative database classified 42 % and 67% of hemiretinas corresponding to non-affected hemifields as abnormal in TD and SD-OCT, respectively (P =0.01). The RNFL measurements were consistently thicker with TD compared with SD -OCT. The RNFL thickness asymetry index in patients with glaucoma was similar with TD [ mean (SD) 0.76 ( 0.17 ) ] and SD-OCT [ mean (SD)0.79(0.12), P = 0,89] and significantly greater than the visual field asymmetry index [ mean (SD ) 0.36 (0.20 ), P < 0.001]. Conclusions: Normal hemifields of glaucoma patients had thinner RNFL measurements than healthy eyes, as measured by TD and SD-OCT. The RNFL measurements were thicker with TD than SD-OCT, SD-OCT detected abnormal RNFL more often than TD-OCT

Yu, Chak Yan Marco.
Thesis (Ph.D.)--Chinese University of Hong Kong, 2013.
Includes bibliographical references (leaves 153-178).
Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web.
Abstract also in Chinese.

Although the current OCT imaging system was designed to examine the retinal structures, a novel application in imaging the anterior chamber angle was studied in section 3.7. OCT was demonstrated to be clinically useful for visualization of the different patterns of angle configurations in different forms of angle closure glaucoma.
In section 3.5, RNFL measurement by OCT was cross-validated by another nerve fiber analyzer, scanning laser polarimetry (SLP). While both OCT and SLP demonstrated comparable diagnostic performance for glaucoma detection and high correlation in the respective RNFL measurements, OCT was found to provide a closer estimation of RNFL thickness with reference to the reported histological measurements. In section 3.6, the structural-functional relationship between RNFL thickness and visual sensitivity was evaluated and compared between OCT and SLP. The relationships were found to be dependent on the choice of the perimetry scale, the type of RNFL measuring devices and the characteristics of the studied subjects. It was concluded that regression analysis of the structural-functional profile could provide important information in the assessment of the trend and pattern of glaucoma progression.
In summary, optical coherence tomography was shown to be useful in the diagnosis of glaucoma and in the evaluation of the trend and pattern of disease progression.
Objectives. The research project was designed to investigate the applications of optical coherence tomography in the assessment of glaucoma. The goals are to identify sensitive and specific anatomic markers, and analytical method for detection of glaucomatous changes, to evaluate the intricate structural-functional relationships in glaucoma with regression analysis and to assess the potential application of optical coherence tomography imaging system in visualization of the anterior chamber angle with a view to obtain OCT data to help understanding the pathophysiology of different forms of angle-closure glaucoma.
Sections 3.1 to 3.3 were designed to identify the most sensitive and specific diagnostic marker(s) for glaucoma detection. Peripapillary retinal nerve fiber layer (RNFL), macular thickness, optic nerve head parameters measured with different reference planes, and a novel anatomic marker - macular nerve fiber layer were investigated. The averaged peripapillary RNFL thickness measured with a high resolution scan (512 scan point) was found to have the best discriminating power for detection of glaucoma. It also has the strongest correlation with visual function. To examine if utilization of the complete data profile of peripapillary RNFL could further improve diagnostic sensitivity, a novel approach with the use of neural network trained to recognize RNFL pattern was studied in section 3.4. It was concluded that neural network analysis could enhance the diagnostic performance for glaucoma detection.
Summary. Glaucoma is a progressive optic neuropathy characterized by the loss of retinal ganglion cells resulting in constriction of visual field and loss of vision as the disease progresses. Since structural damage in glaucoma occurs well before any detectable loss in visual function, clinical examination of the optic nerve head and its nerve fiber layer is crucial in establishing the diagnosis, monitoring the progression and initiating treatment before irreversible damage takes place. The present research project is composed of 7 coherent studies (sections 3.1 to 3.7), aiming to investigate the clinical applications of optical coherence tomography (OCT), an advanced imaging device for detailed examination of optic nerve head and nerve fiber layer, in the assessment of glaucoma.
Leung Kai-shun.
"June 2006."
Adviser: Chi Pui Pang.
Source: Dissertation Abstracts International, Volume: 67-11, Section: B, page: 6323.
Thesis (M.D.)--Chinese University of Hong Kong, 2006.
Includes bibliographical references (p. 212-227).
Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web.
Electronic reproduction. [Ann Arbor, MI] : ProQuest Information and Learning, [200-] System requirements: Adobe Acrobat Reader. Available via World Wide Web.
School code: 1307.

M.Phil.
The record cards of all patients (21 554) examined by the eye clinic on the Phelophepa Mobile Health Care train during January and September, 1995 were analysed. The prevalence of. POAG in the sample was 0-.8% (177 of 21 554 persons). Differences in POAG prevalence were observed with respect to sex, age and geographical region. There was a significant difference (p<0.001) in the number of males (1.1%, 92 of 8113), compared to females (0.6%, 85 of 13 441) diagnosed with POAG. In persons over 40 years of age the prevalence rate was 1.2 % (166 of 14 254 persons) and in persons over 60 years the rate was 1.7% (110 of 6375 persons). The highest prevalence rate was found in the Western Cape (1.84%) and the lowest in the Eastern Cape (0.33%). POAG patients were compared to a control group to investigate the role of certain demographic, systemic and ocular factors. Risk factor analysis identified old age(> 60 years) (OR = 7.2, 95% CI = 4.4 - 11.7), geographical area (Western Cape, OR =2.5, 95% CI = 1.7 - 3.9), systolic hypertension (OR = 2.2, 95% CI = 1.3 - 3.7), diastolic perfusion pressure (< 40 mmHg, OR = 12.9, 95% CI = 4.2 - 52.9), myopia (OR = 2.7, 95% CI = 1.7 - 4.4) and elevated IOP (21 - 30 mmHg, OR = 12.6, 95% CI = 6.4 - 25.0) as significant risk factors. The effectiveness of employing certain blood pressure (BP) and intraocular pressure (TOP) variables as screening tools for glaucoma was evaluated. The systolic BP/IOP ratio was the most valid of the four tests evaluated (sensitivity = 66.0%, specificity = 98.2%, phi coefficient = 0.72). The study recommends that glaucoma screening programmes need to be developed which include sphygmomanometry as part of a battery of tests, and these programmes be targeted specifically at high risk populations (elderly, hypertensives). Further epidemiological studies are required which investigate reasons for the geographical differences found with respect to glaucoma prevalence.

碩士
國立勤益技術學院
工業工程與管理系
93
Traditional medical diagnosis relies on doctors' past experience most. The advanced method of data analysis will enhance the accuracy of disease diagnosis with respect to the multiple examined items. Several researches have already verified that machine learning helps the diagnosis of the disease significantly. This research, based on the newest inspection instrument named Stratus Optical Coherence Tomography (Stratus OCT), is to construct the effective diagnostic system for Taiwanese glaucoma through the application of Back-Propagation Neural Network (BPN) and adaptive neural fuzzy inference system (ANFIS). For both models, we apply Taguchi method on parameter design and variable selection. The diagnosis accuracies are 91.0% for BPN and 89.2% for ANFIS. Though the diagnosis accuracy from BPN is higher than that from ANFIS, the operation process of BPN is like a black box and the readability is lower than ANFIS. We conclude the If-Then rules and the membership functions from ANFIS, and finally ANFIS is suggested to assist the clinical glaucoma diagnosis for early detection.

青光眼作為一種慢性進展性視神經病變，已經成為世界眼科病變中導致不可逆盲的首要原因。青光眼的早期診斷和治療對於降低疾病進展的風險至關重要。光學相干斷層掃描（OCT）可以提供在體視網膜橫斷面的視圖，從而實現了對視網膜神經纖維層（RNFL）改變的客觀測量，這些改變已經被證明了與青光眼引起的視神經損害相關，並已成為診斷青光眼的重要參考依據。
頻域OCT是最新一代的光學相干斷層掃描，它具有比時域OCT更快的掃描速度和更高的圖像解析度，因此，頻域OCT可以提供更可靠的RNFL厚度測量和RNFL缺損評估。本文的研究目的在於評估頻域OCT對RNFL厚度的重測再現性，以及探討影響RNFL厚度測量的因素，這些因素包括（1）影像平均法的應用，（2）RNFL分層錯誤，和（3）視網膜血管的影響。此外，由於RNFL攝影是一個評估青光眼RNFL缺損的臨床參考標準，我們還將其對RNFL缺損的測量與頻域OCT的RNFL厚度偏差圖所作出的測量進行了比較。
首先，為了評估頻域OCT對RNFL厚度測量的重測再現性，15名正常人和15名青光眼患者連續四周每週均接受一次OCT掃描。正常組和青光眼組的RNFL厚度再現性係數分別為4.77-12.65微米和4.53-16.66微米，由於組內相關性係數均大於0.773，說明頻域OCT所作出的RNFL厚度測量是具備可重複性的。
其次，通過分析54隻眼（25名正常志願者和29名青光眼患者）的RNFL厚度測量值，本文對圖像平均法的應用是否會影響RNFL厚度的測量這一問題進行了探討。分析中，每一隻眼均接受了3次OCT掃描，3次掃描的圖像分別使用2、8、和16張連續的圖像進行影像平均。結果顯示，除了青光眼組的鼻下象限RNFL厚度測量值之外（P=0.036），不同的圖像幀數並不會對兩組的總體和其它各象限的RNFL厚度測量值產生顯著的影響（P≥0.055）。雖然圖像平均法的應用對RNFL厚度測量的影響並不顯著，但是視網膜血管和RNFL分層錯誤對青光眼,尤其是對RNFL非常薄的晚期青光眼患者的RNFL厚度測量有影響。結論來自對60個正常人，66個輕至中度青光眼（MD≥-6 dB）患者和54個嚴重青光眼（MD<-6 dB）患者的共180張OCT圖像的分析。視網膜血管相對於平均RNFL厚度的比例均值在正常組，輕至中度青光眼組，和嚴重青光眼組分別為11.2±2.3，12.6±2.5，和16.6±3.9。在人為調整了RNFL界限以糾正RNFL分層錯誤的前後，總體RNFL厚度的差異範圍在正常組為-3.0-2.5微米，輕至中度青光眼組為-2.5-5.0微米，嚴重青光眼組為-11.0-9.5微米組。
最後，通過對41名青光眼患者的51隻眼的RNFL缺損面積，位置，和覆蓋角度進行測量，本文將頻域OCT作出的測量結果和共焦鐳射掃描檢眼鏡（CSLO）RNFL反射影像圖的測量結果進行了比較，結果顯示：OCT不但可以檢測到所有出現在CSLO的RNFL反射影像圖上的RNFL缺損，更重要的是，OCT還可以檢測出額外的並未在RNFL反射影像圖上出現的RNFL缺損。
總之，頻域OCT是一種可提供高再現性RNFL厚度測量的影像方法。對青光眼,尤其是晚期青光眼的RNFL厚度測量值的詮釋，應當考慮到視網膜血管和RNFL分層錯誤的影響。OCT具備對RNFL缺損進行多維度量化（包括厚度，面積，位置，和覆蓋角度）的能力，在青光眼RNFL改變的檢測和監測方面，相對于傳統的RNFL攝影，OCT無疑是更有效的選擇。
Glaucoma, a chronic progressive optic neuropathy, is the leading cause of irreversible blindness in the world. An early diagnosis and treatment of glaucoma is vital to reduce the risk of disease progression. Providing a cross-sectional view of the retina in vivo, optical coherence tomography (OCT) can objectively measure the changes of retinal nerve fiber layer (RNFL), which has been shown to be of relevance and importance in detecting glaucomatous damage of the optic nerve.
The latest generation of OCT, the spectral-domain OCT, has a faster scan speed and a higher image resolution compared to the time-domain OCT. It is expected that the spectral-domain OCT would allow a more reliable measurement of the RNFL thickness and assessment of RNFL defects. The objectives of this research project were to examine the test-retest reproducibility of spectral-domain OCT RNFL measurement and investigate factors including (1) image averaging, (2) segmentation failure, and (3) contribution of retinal blood vessels that might affect the measurement of RNFL thickness. As RNFL photography is a reference standard to evaluate RNFL defects in glaucoma, we also evaluated whether RNFL defects measured in the spectral-domain OCT RNFL thickness map would be comparable to those detected in RNFL photographs.
To evaluate the test-retest reproducibility of RNFL measurements obtained by the spectral-domain OCT, 15 normal individuals and 15 glaucoma patients were followed and imaged weekly for 4 consecutively weeks. The reproducibility coefficients of RNFL thicknesses ranged between 4.53 and 16.66 μm for the normal group, and 4.77 and 12.65 μm for the glaucoma group. The intraclass correlation coefficients were all above 0.773, indicating RNFL measurement with spectral-domain OCT was reproducible.
We then investigated if multiple-image averaging would influence the measurement of RNFL thickness. A total of 54 eyes from 25 normal volunteers and 29 glaucoma patients with RNFL images captured and averaged with 2, 8, and 16 consecutive image frames were analyzed. For both groups, there were no significant differences in global or sectoral RNFL thicknesses among the image series averaged with different number of image frames (all with P≥0.055) except for the inferonasal sector in the glaucoma group (P=0.036). Although the impact of image averaging on RNFL measurement was insignificant, the presence of retinal blood vessels and segmentation errors were influential on the measurement, particularly in advanced glaucoma patients when the RNFL was thin. Analyzing a total of 180 eyes from 60 normal individuals, 66 mild to moderate (MD≥-6 dB) and 54 advanced (MD<-6 dB) glaucoma patients, the mean proportion of retinal blood vessels relative to the average RNFL thickness was 11.2±2.3%, 12.6±2.5% and 16.6±3.9%, respectively. After correcting the segmentation errors by manually refining the RNFL boundaries, the differences in average RNFL thickness ranged from -3.0 to 2.5 m in the normal, -2.5 to 5.0 m in the mild to moderate glaucoma and -11.0 to 9.5 m in the advanced glaucoma groups.
Finally, we compared the area, the angular location, and the angular width of RNFL defects from 51 eyes of 41 glaucoma patients measured with the spectral-domain OCT and RNFL reflectance images obtained by a confocal scanning laser ophthalmoscope (CSLO). OCT was able to detect areas of RNFL abnormalities in all eyes with RNFL defects which were evident in the CSLO RNFL reflectance images. More important, OCT could identify additional RNFL thinning not apparent in RNFL reflectance images.
In summary, spectral-domain OCT could offer an effective approach in measuring RNFL with high reproducibility. Interpretation of RNFL measurement should take the contribution of the retinal blood vessels and segmentation errors into consideration, particularly in advanced glaucoma when the RNFL is thin. With the ability to quantify multiple dimensions of RNFL defects (thickness, area, angular location, and angular width), OCT could provide a useful alternative to detect and monitor RNFL changes in glaucoma.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Ye, Cong.
Thesis (Ph.D.)--Chinese University of Hong Kong, 2012.
Includes bibliographical references (leaves 117-130).
Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web.
Abstract also in Chinese.
ABSTRACT --- p.i
摘要 (ABSTRACT IN CHINESE) --- p.v
DEDICATION --- p.viii
ACKNOWLEDGEMENT --- p.ix
TABLE OF CONTENTS --- p.x
PUBLICATIONS --- p.xiv
ABBREVIATIONS --- p.xvi
Chapter CHAPTER 1: --- INTRODUCTION --- p.1
Chapter 1.1 --- Glaucoma --- p.2
Definition of Glaucoma --- p.2
Epidemiology of Glaucoma --- p.3
Pathogenesis of Glaucoma --- p.4
Diagnosis of Glaucoma --- p.7
Chapter 1.2 --- Retinal Nerve Fiber Layer --- p.13
Anatomy of Retinal Nerve Fiber Layer --- p.13
Visualization of Retinal Nerve Fiber Layer --- p.14
Retinal Nerve Fiber Layer Defect in Glaucoma --- p.16
Significance of Detecting Retinal Nerve Fiber Layer Defect in Glaucoma --- p.18
Chapter 1.3 --- Optical Coherence Tomography --- p.20
Principle of Optical Coherence Tomography --- p.20
Retinal Nerve Fiber Layer Imaging with OCT --- p.21
Optic Nerve Head Imaging with OCT --- p.27
Advantages and Disadvantages of Optical Coherence Tomography --- p.29
Chapter 1.4 --- Research Objectives --- p.30
Chapter CHAPTER 2: --- GENERAL MATERIALS AND METHODS --- p.32
Chapter 2.1 --- Subject Enrollments --- p.33
Chapter 2.2 --- Clinical Ophthalmic Examination --- p.34
Chapter 2.3 --- Visual Field Examination --- p.35
Definition of Normal and Glaucoma Groups --- p.35
Chapter 2.4 --- Optical Coherence Tomography Imaging --- p.37
Cirrus HD-OCT Imaging --- p.37
Spectralis OCT Imaging --- p.37
Chapter 2.5 --- Statistical Analysis --- p.39
Chapter CHAPTER 3: --- RETINAL NERVE FIBER LAYER IMAGING WITH SPECTRAL-DOMAIN OPTICAL COHERENCE TOMOGRAPHY --- p.40
Chapter 3.1 --- Reproducibility and Agreement of Retinal Nerve Fiber Layer Measurement --- p.41
Introduction and Study Objectives --- p.41
Methods --- p.42
Results --- p.45
Discussion --- p.47
Tables and Figures --- p.51
Chapter 3.2 --- Effect of Multiple B-scans Averaging on Retinal Nerve Fiber Layer Measurement --- p.58
Introduction and Study Objectives --- p.58
Methods --- p.59
Results --- p.61
Discussion --- p.62
Tables and Figures --- p.67
Chapter 3.3 --- Impact of Blood Vessels and Segmentation Failure on Retinal Nerve Fiber Layer Measurement --- p.73
Introduction and Study Objectives --- p.73
Methods --- p.75
Results --- p.78
Discussion --- p.80
Tables and Figures --- p.84
Chapter 3.4 --- Agreement of Localized Retinal Nerve Fiber Layer Defect Assessment with Confocal Scanning Laser Ophthalmoscopy --- p.95
Introduction and Study Objectives --- p.95
Methods --- p.97
Results --- p.101
Discussion --- p.103
Tables and Figures --- p.108
Chapter CHAPTER 4: --- GENERAL CONCLUSIONS --- p.115
REFERENCES --- p.117

MMed, Ophthalmology,Faculty of Health Sciences, University of the Witwatersrand, 2009
Objectives To determine in a cohort of 69 African patients with large optic discs and large optic cups, that proportion of patients with physiologic cupping (normal eyes) misdiagnosed as glaucomatous. To evaluate the possible relationship between optic disc size and central corneal thickness. Design and method A case series of 69 patients with large discs (vertical disc height measuring 1.8mm) and large cups (vertical cup to disc ratio 0.6) was evaluated to determine what proportion had glaucoma and what proportion was normal. Patients categorized as normal were further evaluated to determine what proportion were previously misdiagnosed and treated for glaucoma. Patients with a suspected diagnosis of glaucoma, normal tension glaucoma or primary open angle glaucoma were recruited from the glaucoma clinic at St John Eye Hospital. Outcome measures included corrected vertical disc height (VDH), vertical cup to disc ratio (CDR), central corneal thickness (CCT), the relationship between VDH and vertical cup height, the relationship between VDH and CCT, adjusted intraocular pressure (A-IOP), retinal nerve fiber layer analysis and visual fields. vii Results Sixty-nine African patients (138 eyes) with large discs and large cups were evaluated. 41 (59%) were females and 28 (41%) were males. The mean age was 56 years. Of the 69 patients, 51 (74%) had physiologic cupping (normal eyes) and 18 (26%) patients were glaucomatous. Of the group of 51 patients with physiologic cupping, there were 9 patients who were previously misdiagnosed with glaucoma and who had received treatment. VDH ranged between 1.9 and 3.2mm (mean ±SD, 2.3±0.26mm). The distribution analysis of VDH measurements noted the largest cluster around 2.3mm. CCT ranged between 454μm and 618μm (mean±SD, 516±37μm). 107 (77.5%) of the 138 eyes had CCT < 544μm. Conclusion Large cup to disc ratio in relation to large disc size can be normal. It can be misdiagnosed as glaucomatous if objective retinal nerve fiber layer analysis is not carried out. In this study, 9 (18%) patients from a group of 51 patients with physiologic cupping were misdiagnosed as glaucomatous. There was no linear correlation between CCT and VDH in this study. Pearson’s correlation coefficient was 0.13. The majority (77.5%) of eyes had thin corneas (CCT < 544μm).

碩士
國立臺灣海洋大學
資訊工程學系
99
Glaucoma is a serious ocular disease and leads to blindness if it couldn’t be early detected and treated in a proper way. The diagnostic criteria for glaucoma include intraocular pressure measurement, optic nerve head evaluation, retinal nerve fiber layer and visual field defect. The observation of optic nerve head, cup to disc ratio and neural rim configuration are important for early detecting glaucoma in clinical practice. However, the broad range of cup to disc ratio is difficult to identify early changes of optic nerve head. Besides, different ethnic groups possess various features in optic nerve head structures. Therefore, it is still important to develop effective detection techniques to assist clinicians on diagnosing glaucoma at early stages. In this study, the novel idea of integrating image inpainting and active contour model techniques successfully assisted the correct identification of cup and disk regions, ISNT rules, and abnormal distribution of retinal nerve fibers. Fifty two clinical fundus retinal images containing normal and glaucoma cases were applied to the proposed system and the prediction accuracy rate achieves 81% in this study.

碩士
國立勤益技術學院
工業工程與管理系
94
This study is to extract decision rules to differentiate between normal and glaucomatous eyes based on the quantitative assessment of summary data reports from the Stratus optical coherence tomography (OCT) in Taiwan Chinese population. Glaucoma is a kind of association disease. The disease can be affected by many factors, such as inherent genetic or external environment. The Stratus Optical Coherence Tomography (Stratus OCT) consisted of an infrared-sensitive video camera to provide a view of an low-coherence interferometer as light source and detection unit, a video monitor, a computer and an image analysis system. This research tries to combine self-organize map (SOM) and decision tree (DT) to discover hidden inherent rules and set up a distinguishing system of glaucoma. 121 glaucomatous patients and 121 normal patients were included in this study. Twelve decision rules were extracted and the accuracies for training sample and testing sample were 94.6% and 87.1%, respectively. Base on the result, we believe the combination of machine learning can enhance the diagnosis accuracy on medical disease.

Background : Frequency Doubling Perimetry ( FDP ) is a recently developed form of perimetry, which may be more sensitive for detecting visual field loss from glaucoma than conventional Achromatic Automated Perimetry ( AAP ). This thesis was undertaken to study aspects of FDP for the detection of early glaucoma. Method : FDP was compared with other forms of perimetry at one point in time using one dataset ( n = 83 ) and longitudinally over a four - year period with another dataset ( n = 62 ). Several aspects were studied : ( 1 ) the ability of FDP to detect visual field loss earlier than AAP, ( 2 ) its ability to detect early functional abnormality in the presence of mild glaucomatous structural abnormality ( 3 ) visual field topography compared with other perimetry and ( 4 ) its ability to predict future field loss when only the nasal quadrants were considered. Results : When subjects at risk of glaucoma with initial visual field loss on FDP were followed over a three - year period, a significant proportion developed field loss with AAP, whilst those without initial FDP loss did not. FDP detected cases of early glaucomatous optic disc damage, which had not been detected using AAP ; however, there was still a proportion of those with abnormal optic discs which remained normal on FDP. FDP field topography was hill - shaped with the most sensitive point centrally ; however, it was considerably flatter and more sensitive than AAP. Finally, if FDP field loss was only considered significant when it occurred within the nasal step location of the visual field, then this may improve the accuracy of glaucoma diagnosis. Conclusion : This thesis has demonstrated that FDP is not only more sensitive than AAP in the detection of glaucomatous optic disc damage, but it is able to predict future field loss on AAP. FDP may therefore be useful in the early detection and management of glaucoma.
Thesis (Ph.D.)--School of Medicine, 2006.

Glaucoma is characterized by progressive optic nerve head (ONH) deformation and retinal nerve fiber layer (RNFL) thinning but the relative sequence of ONH and RNFL changes in glaucoma remains largely uncertain. It has been proposed that structural damage of the optic nerve can often be detected before detectable functional loss. Therefore, investigating structural changes of the ONH and RNFL is of importance and relevance in the monitoring and management of glaucoma progression. Spectral domain optical coherence tomography (OCT), scanning laser polarimetry (SLP) and confocal scanning laser ophthalmoscopy (CSLO) are the three prevailing digital imaging technologies for measurement of RNFL thickness, RNFL retardance and ONH parameters, respectively. Although these instruments have been extensively investigated for detection of glaucomatous damage, less is known about their relative performance for detection of change in glaucoma progression. Although previous studies on non- human primates showed that disruption of the microtubule structure of the retinal ganglion cell axons detected by SLP as reduction of RNFL retardance, as well as ONH surface deformation detected by CSLO, could be detected prior to reduction of RNFL thickness measured with OCT, clinical data corroborating this observation are lacking. The sequence of change of RNFL thickness, RNFL retardance and ONH parameters has not been investigated in human glaucoma.
This research project aimed to investigate the performance of OCT, SLP and CLSO for change detection of RNFL and ONH damages, determine the relative sequence of change of RNFL retardance and RNFL thickness and ONH deformation, and evaluate if ocular biomechanical properties, measured as corneal hysteresis by the ocular response analyzer (ORA, Reichert Inc.), influence the detection of ONH and RNFL progression in glaucoma patients. We hypothesized that ONH deformation and loss of RNFL retardance could be detected before detectable RNFL thinning and that the baseline corneal hysteresis would be a risk factor for ONH and RNFL damage in glaucoma.
In the first study, we analyzed 184 eyes of 116 patients with glaucoma and 43 normal eyes of 23 healthy individuals followed for a mean of 4.6 years. All subjects had RNFL retardance and RNFL thickness measurements obtained with GDx ECC (Carl Zeiss Meditec) and Cirrus HD-OCT (Carl Zeiss Meditec), respectively, at 4-month intervals. Progressive reduction of RNFL retardance and RNFL thickness were evaluated with Guided Progression Analysis (GPA, Carl Zeiss Meditec) with reference to the RNFL retardance change map and RNFL thickness change map, respectively. Twenty seven eyes of 26 patients showed progressive RNFL thinning whereas 8 eyes of 8 patients had RNFL retardance reduction in the latest follow-up visit. Seven eyes of 7 patients had progressive RNFL thinning and reduction of RNFL retardance detected by both instruments; all had progressive RNFL thinning evident before reduction of RNFL retardance and the mean lag time was 13.4 months (range: 4.0-37.6 months). The survival time of eyes detected with RNFL thinning was significantly shorter than the survival time of eyes detected with reduction of RNFL retardance (P=0.001). No eyes in the normal group showed progressive RNFL changes during follow-up. Collectively, we showed that at a comparable specificity (100%, 95% confidence interval: 96.3%-100%), progressive RNFL thinning was detected more often than progressive reduction of RNFL retardance and the former preceded the latter in eyes with both progressive RNFL thinning and reduction of RNFL retardance.
In the second study using a similar study design, we investigated the sequence of change of ONH surface depression detected by CSLO (HRT 3, Heidelberg Engineering) and RNFL thinning detected by OCT (Cirrus HD-OCT, Carl Zeiss Meditec) in 146 eyes of 90 glaucoma patients followed at approximately 4-month intervals for an average of 5.4 years. Significant ONH surface depression and RNFL thinning were defined with reference to Topographic Change Analysis (TCA, Heidelberg Engineering) and Guided Progression Analysis (GPA, Carl Zeiss Meditec), respectively. At a specificity of 94.3% (95% confidence interval: 86.2%-97.8%) for both RNFL thinning and ONH surface depression (determined in a normal group comprising 70 eyes from 35 normal subjects), 57 eyes (39.0%) had ONH surface depression, 46 eyes (31.5%) had RNFL thinning, and 23 eyes (15.8%) had both in the glaucoma group. Among these 23 eyes, 19 (82.6%) had ONH surface depression detected prior to RNFL thinning and the median lag time was 15.8 months (range, 4.0-40.8 months). Although only 7.0% of eyes (4/57) had RNFL thinning at the onset of ONH surface depression, 45.7% (21/46) had ONH surface depression at the onset of RNFL thinning. The survival time of eyes with ONH surface depression was significantly shorter than the survival time of eyes detected with RNFL thinning (P=0.002). With reference to the HRT TCA and OCT GPA, ONH surface depression occurred before RNFL thinning in a significant proportion of patients with glaucoma at a comparable specificity.
Of note, a significant proportion of eyes had ONH surface depression without any detectable progressive RNFL thinning in the second study, and vice versa. Investigating whether the risk factors for ONH surface depression and RNFL progression are different is therefore important. In the final study, we investigated if baseline corneal hysteresis is a risk factor for progressive ONH surface depression and RNFL thinning. Following the same cohort of 146 eyes of 90 glaucoma patients for an average of 6.8 years, we detected that 65 eyes (44.5%) had progressive ONH surface depression, 55 eyes (37.7%) had progressive RNFL thinning and 20 eyes (13.7%) had visual field progression (based on the EMGT criteria). After adjusting for ages, CCT, baseline diastolic IOP, average IOP during follow-up, baseline disc area and baseline MD in the cox proportional hazards model, baseline corneal hysteresis was significantly associated with ONH surface depression (HR=0.70, P=0.008), visual field progression (HR=0.56, P=0.019), but not with progressive RNFL thinning (HR=0.96, P=0.751). For each 1-mmHg decrease of baseline CH, the hazards for ONH surface depression and visual field progression increased by 30% and 44%, respectively.
In summary, at a comparable level of specificity, progressive ONH surface depression detected by CSLO could be observed prior to progressive RNFL thinning detected by OCT, which preceded identified reduction of RNFL retardance detected by SLP. For eyes with concomitant ONH surface depression, RNFL thinning and visual field progression, ONH surface depression always preceded visual field progression. Our finding indicates that a time window for therapeutic intervention may exist upon detection of ONH surface depression before irreversible RNFL and visual field loss and that measurement of CH would be useful to predict ONH surface depression and visual field progression.
Further studies are required to investigate the sequence of optic nerve head change and RNFL progression with the same instrument. Whether IOP lowering treatment initiated at the time of ONH deformation would be effective to prevent or slow down RNFL and visual field loss needs to be further investigated. A more reliable and accurate measure of the ocular biomechanical properties is necessary for evaluation of their contribution to glaucoma progression.
青光眼是一種進展性視神經病變，其特徵為﹕視神經乳頭變形，神經纖維層（RNFL）的變薄以及相應的視野缺損。然而，青光眼結構性改變和功能性變化發生的相對順序仍不清楚。視神經結構性改變被認為要早於功能性改變的發生。因此，研究視乳頭的結構性改變具有重要意義，有助於早期診斷青光眼的進展及隨訪青光眼患者。目前主要用於RNFL厚度，RNFL阻滯性以及視乳頭參數的影像學掃描儀器為頻域OCT，鐳射偏振光掃描器（SLP）和共聚焦鐳射掃描眼底鏡（CSLO）。儘管這三種儀器已經廣泛用於青光眼損傷的檢測，但在青光眼患者結構性變化的應用並不常見。既往在非人靈長類動物的實驗中，通過破壞神經節細胞軸突中的微小管結構，從而發現RNFL的阻滯性以及視乳頭的變化要先於RNFL厚度變化的發生。然而在臨床研究中並未得到證實。同時，在青光眼患者中，RNFL厚度變化，RNFL阻滯減少以及視神經頭參數改變之間的先後順序並未得到證實。
本次實驗研究的目的在於探討OCT，SLP及CSLO在診斷青光眼病人RNFL及視乳頭進展的能力，確定RNFL厚度變化，RNFL阻滯性減少以及視神經頭參數改變之間的相對順序，以及評估眼反應分析儀（ORA）測得的角膜粘滯性（CH）是否影響視乳頭及RNFL厚度的進展。我們假設：視神經乳頭的變形，RNFL阻滯性的減少要先於RNFL厚度的變化，基線角膜粘滯性的測量會影響視乳頭及RNFL進展的檢測。116個青光眼病人的184隻眼以及23個正常對照的43隻眼被納入第一個研究中。所有受試物件均接受每4個月一次的OCT以及SLP RNFL的掃描，平均隨訪時間為4.6年。通過OCT及SLP中Guided Progression Analysis（GPA, Carl Zeiss Meditec）程式，一系列RNFL厚度及粘滯性圖被自動分析從而獲得RNFL厚度及粘滯性的變化結果。26個青光眼患者的27隻眼表現為RNFL厚度的進行性變薄，8個患者的8隻眼表現為RNFL粘滯性的減少。其中7個患者的7隻眼同時表現為RNFL厚度變薄及粘滯性的減少，所有這7隻眼的RNFL變薄的發生要早於RNFL粘滯性的減少，兩者間隔時間平均為13.4月（4.0-37.6月）。RNFL厚度變薄者的生存概率明顯小於RNFL粘滯性減少的青光眼患者（P=0.001）。隨訪中，我們未發現正常對照組中RNFL厚度變薄或者粘滯性改變者。總體說來，在同一特異性水準（100%），RNFL厚度的變化頻率高於粘滯性的改變，RNFL厚度的變薄要早於粘滯性減少的發生。
採用相同於第一個研究的研究方法，我們研究CSLO測得的視乳頭表面凹陷以及測得OCT的RNFL厚度變化發生的相對順序。90個青光眼患者的146隻眼以及35個正常對照物件的70隻眼被納入第二個研究中。所有受試物件均接受4個月一次的CSLO及OCT掃描從而獲得一系列的視神經頭表面的拓撲圖像以及RNFL厚度圖。CSLO TCA及OCT GPA程式自動對比基線及隨訪中所獲得的視神經頭表面的拓撲圖像以及RNFL厚度圖，從而獲得視乳頭表面凹陷及RFNL進展報告。平均隨訪5.4年後，CSLO及OCT在診斷視神經頭及RNFL進展的特異性為94.3%，57只青光眼患眼（39.0%）表現為顯著性視乳頭表面凹陷，46隻眼（31.5%）表現為RFNL厚度的進行性變薄，而23隻眼（15.8%）同時表現為視乳頭面凹陷和RFNL的進行性變薄。在這23只眼中，19隻眼（82.6%）變現為視乳頭表面凹陷先於RFNL厚度變薄的發生，間隔時間的中值為15.8個月（4.0-40.8月）。儘管在顯著性視乳頭表面凹陷發生時，僅有7.0%的患眼表現為RNFL厚度的變薄；但是，在RNFL厚度發生顯著性變薄時已有45.7%的患眼表現為視乳頭表面凹陷。視乳頭表面凹陷患眼的生存概率差於RNFL厚度變薄患眼（P=0.002）。在青光眼患者的隨訪中，CLSO TCA測得的視乳頭表面凹陷要早於OCT GPA測得的RNFL厚度的變化。
最後的一個研究目在於評估眼反應分析儀（ORA）測得的基線角膜粘滯性（CH）是否為視乳頭表面凹陷及RNFL厚度變薄的危險因素。平均隨訪同一人群即第二個研究中的90個青光眼患者的146眼6.8年，65隻眼（44.5%）被檢測出具有進行性視乳頭表面凹陷，55隻眼（37.7%）表現為進行性RNFL厚度的變薄，20隻眼（13.7%）表現為進行性視野的缺損（基於EMGT標準）。基線CH與視乳頭表面凹陷，視野進展間具有顯著性相關關係（HR=0.70，P=0.008及HR=0.56，P=0.019），但CH與進行性RNFL厚度變薄間並無顯著性相關關係（HR=0.96，P=0.751）。每1毫米汞柱基線CH的降低，發生視乳頭表面凹陷及視野缺損的危險性將增加30%及44%。CH的測量值與青光眼進展的危險性具有顯著相關關係。
總之，在具有可比性特異性水準下，CSLO檢測的進展性視乳頭表面凹陷的發生要先於OCT檢測的進行性RNFL厚度的變薄，後者的發生早於SLP測得的RNFL粘滯性的改變。對於同時有視乳頭表面凹陷，RNFL厚度變薄及RNFL粘滯性改變的青光眼患眼，視乳頭表面凹陷的發生要早於視野的進展。我們的實驗研究表明了在青光眼患者發生視乳頭表面凹陷時，治療的時間窗的存在有助於避免不可逆的RNFL缺失及視野的缺損。角膜粘滯性的測量對於預測視乳頭表面凹陷及視野進展具有重要意義。
展望未來的研究中，用同一種儀器進行視乳頭及神經纖維層的隨訪，從而得出相對的變化次序很有必要。研究在視乳頭或者神經纖維層發生變化時進行眼壓的干預是否能避免視功能的進一步損傷顯得尤為重要。用於測量角膜生物學特性的更為準確，可信度更高的儀器真正研發中，以及進一步探討角膜生物學特性與青光眼進展之間的關係。
Xu, Guihua.
Thesis Ph.D. Chinese University of Hong Kong 2015.
Includes bibliographical references (leaves 116-145).
Abstracts also in Chinese.
Title from PDF title page (viewed on 18, October, 2016).
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.

M.Phil. (Optometry)
The primary aim of this research was to investigate whether the Icare® rebound tonometer may be used in place of the Goldmann tonometer to obtain accurate and reliable intraocular pressure measurements on a sample of the general population seeking eye care, in a South African context. Due to the portability of the Icare®, lack of dependency on other instrumentation and power source, together with the ease of use with minimal training, and without the use of topical anaesthetic favourable results of the instrument may lead to its widespread use. This could aid in earlier diagnosis of glaucoma where intraocular pressure remains the only modifiable risk factor. In the South African context, where a high prevalence of undiagnosed primary open angle glaucoma has been found, access and accuracy of intraocular pressure measurement could save and extend functional vision in this country. Patients presenting at the Department of Optometry for routine eye examinations were selected utilizing convenience sampling. The right and left eyes of 113 patients were assessed using both the Icare® TA01i and the Goldman applanation tonometer by independent examiners for each method, all readings of intraocular pressure with the Goldman being taken by the same experienced examiner. The age of the patients ranged from 20 to 89 years with a mean age 50.29 ±20.97 years. Using the Oculus Pachycam® when it became available, central corneal thickness was obtained on 71 patients (142 eyes). Analysis of data using descriptive statistics from SPSS (Statistical Programs for the Social Sciences) was performed in addition to the Bland-Altman method of comparative analysis for sets of data of corrected and uncorrected measurements between the instruments...

碩士
淡江大學
數學學系碩士班
99
Meta-analysis is a quantitative weighted average method to combine the results of related but independent studies (usually drawn from the published literatures) and synthesize summaries and conclusions which may be used to evaluate the therapeutic effects and/or plain new study accordingly. The commonly used meta-analyses, dependent on the characteristic of data, are: the relative risk, odds ratio, and rates difference for binary data and effect size and meta-regression for normally distributed data. Meta-analysis for area under ROC curve (AUC), a commonly used medical diagnosis method, has not been proposed yet. Glaucoma is an irreversible optic neuropathy, which characterized by progressive retinal nerve fiber layer (RNFL) thinning. Its severity could be evaluated objectively by imaging techniques which mainly by optical coherence tomography (OCT) in current ophthalmology. The diagnostic capacities of OCT for glaucoma were heterogeneous in the current published literatures. Most of them were presented in terms of sensitivity, specificity, and area under ROC curve (AUC). Accordingly, a random effects’ meta-analysis for AUC will be helpful to synthesize the overall diagnostic capacities of OCT for glaucoma. In this study, we are going to propose a fixed/random effects meta-analysis method for area under ROC curve and apply it to glaucoma diagnostic by OCT.

博士
國立陽明大學
公共衛生研究所
103
Purpose: The study compared the general and ocular biometric characteristics of different stages and different subtypes of primary angle-closure glaucoma. In the first section of the study, we compared the general and ocular biometric characteristic of control, primary angle closure (PAC), and primary open-angle glaucoma (PACG) patients to better understand the possible relationship between differences in ocular parameters that might predict risk for PACG in PAC patients. In the second section of our study, we compared the ocular biometrical parameters of normal, plateau iris, and pupillary block eyes by A-scan ultrasonography to distinguish the frequently misdiagnosed plateau iris eyes from pupillary block group and normal group. Methods: In the first section of our study, one hundred normal, 90 PAC, and 90 PACG eyes were retrospectively reviewed. In the second section of our study, we retrospectively reviewed 71 normal, 39 plateau iris, and 83 pupillary block eyes. General characteristics such as age, gender, body height, body weight, blood pressure, pulse, systemic diseases, and education level were recorded. Ocular findings included visual acuity, intraocular pressure, refraction, cup to disc ratio, and ocular biometry. Ocular biometry was obtained by A-scan ultrasonography (Digital A/B scan 5500; Sonomed Inc.). The parameters recorded were anterior chamber depth (ACD), lens thickness (LT), axial length (AXL), lens/axial length factor (LAF), and relative lens position (RLP). Results: In the first section of our study, the controls, PAC group, and PACG group were found to be significantly different in age (62.7±9.8; 65.3±7.5; and 66.0±7.4, respectively). With regard to ocular parameters, the ACD tended to decrease and the LT and LAF tended to increase from normal to PAC to PACG. The eyes of the PACG group had significantly shallower ACD (P<0.001) and thicker lens (P<0.001) than those of the PAC group. While PAC had similar lens position to the control group, PACG had more anteriorly positioned lens than the PAC group (P<0.001). Logistic regression analysis found a significant association between a decrease in ACD and increased risk of PACG (odds ratio=3.59 for 0.2mm decrease in ACD) as well as a significant association between an increase in LT and increased risk of PACG (odds ratio=1.30 for 0.2mm increase in LT). In the second section of our study, the normal controls, plateau iris group and pupillary block group were significantly different in age, but not in gender. The anterior chamber depth tended to decrease and the lens thickness tended to increase from normal to plateau iris to pupillary block eyes. Compared to those of plateau iris group, the pupillary block group had significantly shallower anterior chamber depth (2.90mmvs. 2.33mm; p<0.001), thicker lens (4.77mmvs. 5.11mm; p<0.001), shorter axial length (23.16mmvs. 22.63mm; p<0.001), smaller relative lens position (2.28 vs. 2.16; p<0.001) and larger lens/axial length factor (2.06 vs. 2.26; p<0.001). However, when comparing plateau iris and normal eyes, only axial length and lens/axial length factor were significantly different (23.16mm vs. 23.54mm; p<0.05 and 2.06 vs. 1.96; p<0.05). Conclusions: In the first section of our study, we found that in addition to LT, a shallower ACD owing to a change in RLP may have a role in the progression from PAC to PACG. Owing to the differences of certain biometric characteristics between PAC and PACG, A-scan ultrasonography might potentially be used for the early detection of PACG in PAC eyes. In the second section of our study, the ocular biometrics of plateau iris measured by A-scan ultrasonography were significantly different from those of pupillary block eyes. However, our A-scan ultrasongraphy generally found no significant biometric differences between plateau iris and normal eyes. These findings suggest that while A-scan ultrasonography might be used as a practical tool for differentiating plateau iris and papillary block eyes, a more meticulous gonioscopy and other assessments may be necessary to distinguish plateau iris from normal eyes.

碩士
中原大學
醫學工程研究所
96
Abstract Glaucoma is one of familiar blind diseases in the ophthalmology. Treatments only can temporarily decelerate worse degree, it’s important to discover and treat in early stages. Optical coherence tomography (OCT) is the major method for the early detection of Glaucoma. However, only depending on OCT may induce optical disc positioning error and reduce the accuracy of diagnosis. The purpose of this study is focused on disc position detection and feature analysis for glaucoma diagnosis in integrating OCT and Fundus image. In the study, Fundus image and OCT disc photo, disc scans in six directional angles were used for Quantitative analysis. First, disc and cup in color Fundus images are segmented by semi-automatic method, following matching disc and cupping with OCT disc photo, getting correct scan center, further against OCT disc scans of the six directional angles. Image parameters which include diameter and area of disc, cup and rim, cup/disc (C/D) ratio, and rim volume were then calculated and analyzed. The shape, boundary, and serious degree of disc and cup were found during this procedure. Totally, 4 Normal cases and 10 glaucoma cases were used for test. First of all, normal cases are evaluated in Quantitative analysis and then evaluated the Glaucoma cases. Finally, the Correlation analysis between computer generated segmentation and ophthalmologist’s manual segmentation was used to evaluate this system. This study finds that the position of scan center and the accuracy of cup boundary segmentation will affect the calculation of Rim volume and correlation coefficient. In order to get correct scan center, we revise OCT scan center 15 pixels by statistics. Based on temporal distance compression ratio of eye, the distances from center to disc and cup was correct. Result showed that average correlation coefficient in normal cases is 0.85, average C/D ratio is 0.86, and Rim volume is 0.59. For glaucoma cases, average correlation coefficient is 0.86, average C/D ratio is 0.88, and Rim volume is 0.59, respectively. In conclusion, this study provide an OCT computer-aided diagnosis interface, let ophthalmologist or user operate conveniently. This system also solve optical disc and cup positioning error in Ophthalmology OCT computer-aided diagnosis. However, system limited to that disc or cup boundary is fuzzy, and disc blood vessel surroundings are bleeding. In addition, Rim volume is not only major parameter in diagnosis of glaucoma. If OCT retinal fiber layer (RNFL) image can be used to evaluate parameter in each angles, the accuracy can be raised.

碩士
淡江大學
數學學系碩士班
99
Meta-analysis is a quantitative weighted average method to combine the results of related but independent studies (usually drawn from the published literatures) and synthesize summaries and conclusions which may be used to evaluate the therapeutic effects and/or plain new study accordingly. The commonly used meta-analyses, dependent on the characteristic of data, are: the relative risk, odds ratio, and rates difference for binary data and effect size and meta-regression for normally distributed data. Meta-analysis for area under ROC curve (AUC), a commonly used medical diagnosis method, has not been proposed yet. Glaucomais an irreversible optic neuropathy, which characterized by progressive retinal nerve fiber layer(RNFL)thinning.Its severity could be evaluated objectively by imaging techniques which mainly by optical coherence tomography (OCT) in current ophthalmology. However, subject’s ethnicity and age are considered as clinically important factors for RNFL measurements. The diagnostic capacities of OCT for glaucoma were heterogeneous in the current published literatures. Most of them were presented in terms of sensitivity, specificity, and area under ROC curve (AUC). Accordingly, meta regression analysiscan be used to explore the potential prognostic factors for the heterogeneityof AUC that will be helpful to synthesize the overall diagnostic capacities of OCT for glaucoma.