Eye Fundus Image Analysis for Automatic Detection of Diabetic Retinopathy

Diabetes is a rapidly increasing worldwide problem which is characterised by defective metabolism of glucose that causes long-term dysfunction and failure of various organs. The most common complication of diabetes is diabetic retinopathy (DR), which is one of the primary causes of blindness and visual impairment in adults. The rapid increase of diabetes pushes the limits of the current DR screening capabilities for which the digital imaging of the eye fundus (retinal imaging), and automatic or semi-automatic image analysis algorithms provide a potential solution. In this work, the use of colour in the detection of diabetic retinopathy is statistically studied using a supervised algorithm based on one-class classification and Gaussian mixture model estimation. The presented algorithm distinguishes a certain diabetic lesion type from all other possible objects in eye fundus images by only estimating the probability density function of that certain lesion type. For the training and ground truth estimation, the algorithm combines manual annotations of several experts for which the best practices were experimentally selected. By assessing the algorithm’s performance while conducting experiments with the colour space selection, both illuminance and colour correction, and background class information, the use of colour in the detection of diabetic retinopathy was quantitatively evaluated. Another contribution of this work is the benchmarking framework for eye fundus image analysis algorithms needed for the development of the automatic DR detection algorithms. The benchmarking framework provides guidelines on how to construct a benchmarking database that comprises true patient images, ground truth, and an evaluation protocol. The evaluation is based on the standard receiver operating characteristics analysis and it follows the medical practice in the decision making providing protocols for image- and pixel-based evaluations. During the work, two public medical image databases with ground truth were published: DIARETDB0 and DIARETDB1. The framework, DR databases and the final algorithm, are made public in the web to set the baseline results for automatic detection of diabetic retinopathy. Although deviating from the general context of the thesis, a simple and effective optic disc localisation method is presented. The optic disc localisation is discussed, since normal eye fundus structures are fundamental in the characterisation of DR.

Meio semi-seletivo para detectar Xanthomonas axonopodis pv. phaseoli em sementes de feijoeiro e sua erradicação através do tratamento de sementes com o fungicida tolylfluanid

Estudou-se a eficiência do meio semi-seletivo desenvolvido para Xanthomonas axonopodis pv. malvacearum (Xam) do algodoeiro, com algumas modificações, na detecção de Xanthomonas axonopodis pv. phaseoli (Xap) em sementes de feijoeiro, bem como a eficiência do fungicida tolylfluanid na erradicação de Xap através do tratamento de sementes. Foram utilizadas sementes naturalmente infectadas por Xap, procedentes de diferentes municípios do Estado do Paraná. Houve o desenvolvimento de colônias bacterianas ao redor das sementes infectadas sob meio semi-seletivo entre seis e 12 dias de incubação. A freqüência de recuperação de colônias da bactéria em relação ao meio Agar nutriente (AN) variou entre 30 a 112%. O crescimento de Xap foi inibido a concentrações entre 54 a 1500 ppm do fungicida tolylfluanid. A bactéria foi recuperada das sementes não tratadas em meio semi-seletivo, mas não das sementes tratadas em solução do fungicida tolylfluanid (1,20g/L água). Em casa de vegetação, das 400 sementes tratadas com tolylfluanid não foi produzida nenhuma planta com sintomas da doença até 30 dias após a semeadura, enquanto que 9,75% das sementes não tratadas apresentaram sintomas de Xap. Em um outro experimento, das 1200 sementes não tratadas um total de 7,08% e 11,67% das plantas mostraram-se com sintomas da doença, enquanto que das sementes tratadas 0,5% e 2,4% apresentaram sintomas, 26 e 46 dias após a semeadura, respectivamente.

Semi-selective medium for Fusarium graminearum detection in seed samples

Fungi of the genus Fusarium cause a variety of difficult to control diseases in different crops, including winter cereals and maize. Among the species of this genus Fusarium graminearum deserves attention. The aim of this work was to develop a semi-selective medium to study this fungus. In several experiments, substrates for fungal growth were tested, including fungicides and antibiotics such as iprodiona, nystatin and triadimenol, and the antibacterial agents streptomycin and neomycin sulfate. Five seed samples of wheat, barley, oat, black beans and soybeans for F. graminearum detection by using the media Nash and Snyder agar (NSA), Segalin & Reis agar (SRA) and one-quarter dextrose agar (1/4PDA; potato 50g; dextrose 5g and agar 20g), either unsupplemented or supplemented with various concentrations of the antimicrobial agents cited above. The selected components and concentrations (g.L-1) of the proposed medium, Segalin & Reis agar (SRA-FG), were: iprodiona 0.05; nystatin 0,025; triadimenol 0.015; neomycin sulfate 0.05; and streptomycin sulfate, 0.3 added of ¼ potato sucrose agar. In the isolation from seeds of cited plant species, the sensitivity of this medium was similar to that of NSA but with de advantage of maintaining the colony morphological aspects similar to those observed in potato-dextrose-agar medium.

Semi-selective culture medium for Exserohilum turcicum isolation from corn seeds

Northern corn leaf blight, caused by Exserohilum turcicum (Et), is a disease of widespread occurrence in regions where corn, sweetcorn and popcorn are grown. This disease has great potential to cause damage and has been studied for years, but the association of its causal agent with seeds remains unconfirmed. Thus, the availability of a sensitive method to detect and quantify the inoculum in seeds, even at low incidence, is essential. The aim of this study was to develop a method to detect and quantify the presence of the fungus infecting and infesting corn and popcorn seeds. Artificially and naturally infected seeds were employed to develop the medium. The semi-selective medium was composed of carbendazim (active ingredient) (60 mg/L), captan (30 mg/L), streptomycin sulfate (500 mg/L) and neomycin sulfate (600 mg/L) aggregated to the medium lactose casein hydrolysate agar medium. By using this, Et was detected in naturally infected corn seeds, showing 0.124% incidence, in four out of ten analyzed samples. In addition, 1.04 conidia were detected per infested seed. By means of isolation, pathogenicity test, morphological characterization and comparison with descriptions of the species in the literature, the fungus isolated from the seeds was confirmed to be Et. Both infection and infestation were considered low; thus, for studies of Et detection in corn seeds, the use of semi-selective medium and more than 1,200 seeds/sample is suggested.

Meios de cultura semi-seletivos para Macrophomina phaseolina

Macrophomina phaseolina (Tassi) Golidanich é um fungo habitante do solo importante economicamente devido ao amplo número de espécies de plantas que infectam e da dificuldade do seu controle. Vários estudos envolvendo densidade de inóculo, taxonomia, sobrevivência, necessitam de meios de cultura seletivo ou semi-seletivo. O objetivo do trabalho foi avaliar 16 meios de cultura quanto à especificidade a este patógeno, proporcionando maior porcentagem de detecção do seu crescimento e menor número de contaminações, para substituir o meio semi-seletivo RB modificado, rotineiramente utilizado em estudos deste patógeno. O meio semi-seletivo RB modificado é bastante eficiente e contém em sua composição o fungicida metalaxyl (inibidor de oomycetos), que atualmente não se encontra disponível comercialmente em formulação simples, sem adição do Mancozeb ou Clorotalonil que inibem o crescimento do fungo M. phaseolina. Os meios de cultura avaliados foram repicados com o inóculo do fungo produzido em substrato areno-orgânico, contido em bolsas de náilon, recuperados após 30 dias de um solo não autoclavado, contido em uma bandeja. Cada meio de cultura avaliado tiveram 7 repetições, representadas por uma placa de Petri. Para as comparações das médias das porcentagens do crescimento de M. phaseolina e do número de contaminantes foi utilizado o teste de Scott-knott a 5% de probabilidade e os valores em porcentagem foram transformados em arc sem (√/100). Dentre os meios de cultura avaliados os MSTP 1 [(BDA com tetraciclina 50 mg.L-1 mais propamocarb a 1 mL.L-1(Previcur N® 72,2% p.a.)], MSRP 0,5 (BDA com rifampicina 100 mg.L-1 mais fungicida propamocarb a 0,5 mL.L-1) e MSRP 1 (BDA com rifampicina 100 mg.L-1 mais fungicida propamocarb a 1 mL.L-1) proporcionaram maior porcentagem e detecção do fungo M. phaseolina e menor número de contaminações por outros fungos e bactérias. Estes meios de cultura semi-seletivos podem ser utilizados em futuros trabalhos visando estudos epidemiológicos e de medidas de controle do fungo M. phaseolina.

Biocontrol of Sclerotinia sclerotiorum and white mold of soybean using saprobic fungi from semi-arid areas of Northeastern Brazil

ABSTRACTThe incidence and the levels of yield loss caused by the white mold of soybean (caused by the fungus Sclerotinia sclerotiorum) have increased in areas of higher altitude at Cerrado and Southern Brazil, causing yield losses of up to 60%. The aim of this study was to select saprobic fungi with the potential to control the white mold of soybean. First, in vitroantagonism screening was carried out to test eight saprobic fungi against S. sclerotiorum. Assessment of S. sclerotiorum mycelial growth was done at four and seven days after its placement on the culture medium. The isolate showing greatest antagonistic effect in all tests/assessments was Myrothecium sp. An in vivo experiment was conducted in a greenhouse and growth chamber, where plants previously treated with eight saprobic fungi were artificially inoculated with S. sclerotiorum. The fungal culture medium (potato-dextrose) and the commercial resistance inducer acibenzolar-S-methyl were used as controls. In the in vivotests, severity of the white mold was assessed at 8, 14 and 21 days after inoculation. The highest reduction percentage in the lesion length was observed for the treatment with Myrothecium sp. (70%), which has the greater potential to be used as biocontrol agent of soybean under the conditions of this experiment.

Estrutura genética espacial de duas populações naturais de Myracrodruon urundeuva M. Allemão na região semi-árida, Brasil

Para determinar se a distribuição espacial dos genótipos é aleatória ou se está estruturada, utilizou-se a autocorrelação espacial dos genótipos para investigar dados alozímicos de duas populações naturais de Myracrodruon urundeuva do semi-árido brasileiro (Estação Ecológica do Seridó/RN, com 1.166,38/ha e Sítio Mata dos Alves/PB com 84/ha). A primeira população encontra-se em áreas mais preservadas, enquanto a segunda está em uma área fragmentada. O programa utilizado foi o "Autocorr". A autocorrelação estimada para os locos polimórficos foi realizada nos indivíduos adultos. Foram analisados dois alelos na Estação Ecológica do Seridó e quatro no Sítio Mata dos Alves. Utilizaram-se de três métodos para o pareamento de indivíduos a serem comparados: conexão de Gabriel (Ig), vizinho mais próximo (Ivmp) e comparações dentro de classes de distâncias preestabelecidas. Os resultados obtidos para Ig e Ivmp (Pgm-2: 0,032 e 0,236; Est-1: 0,263 e 0,242, respectivamente) na estação ecológica e Ig e Ivmp (Pgm-1: -0,349 e -0,288; Pgm-2: -0,341 e-0,278; Est-1: -0,349 e -0,284 e Mdh-1: -0,345 e 0,282, respectivamente) no sítio não mostraram a presença de estruturação genética espacial, o que possibilita pressupor que exista uma distribuição aleatória dos genótipos dentro delas. O mesmo foi detectado para as comparações dentro de classes de distâncias preestabelecidas. Os resultados mantiveram o mesmo padrão encontrado para algumas populações naturais de espécies arbóreas tropicais já estudadas.

Chlorophyll fluorescence of tropical tree species in a semi-deciduous forest gap

The characterization of different ecological groups in a forest formation/succession is unclear. To better define the different successional classes, we have to consider ecophysiological aspects, such as the capacity to use or dissipate the light energy available. The main objective of this work was to assess the chlorophyll fluorescence emission of tropical tree species growing in a gap of a semi-deciduous forest. Three species of different ecological groups were selected: Croton floribundus Spreng. (pioneer, P), Astronium graveolens Jacq. (early secondary, Si), and Esenbeckia febrifuga A. Juss. (late secondary, St). The potential (Fv/Fm) and effective (deltaF/Fm') quantum efficiency of photosystem II, apparent electron transport rate (ETR), non-photochemical (qN) and photochemical (qP) quenching of fluorescence were evaluated, using a modulated fluorometer, between 7:30 and 11:00 h. Values of Fv/Fm remained constant in St, decreasing in P and Si after 9:30 h, indicating the occurrence of photoinhibition. Concerning the measurements taken under light conditions (deltaF/Fm', ETR, qP and qN), P and Si showed better photochemical performance, i.e., values of deltaF/Fm', ETR and qP were higher than St when light intensity was increased. Values of qN indicated that P and Si had an increasing tendency of dissipating the excess of energy absorbed by the leaf, whereas the opposite was found for St. The principal component analysis (PCA), considering all evaluated parameters, showed a clear distinction between St, P and Si, with P and Si being closer. The PCA results suggest that chlorophyll fluorescence may be a potential tool to differentiate tree species from distinct successional groups.

Resistência natural de nove madeiras do semi-árido brasileiro a fungos xilófagos em condições de laboratório

Objetivou-se avaliar a resistência natural de nove madeiras do semi-árido brasileiro a fungos xilófagos, em condições de laboratório. As madeiras estudadas foram a algaroba (Prosopis juliflora), o angico (Anadenanthera macrocarpa), a aroeira (Myracrodruon urundeuva), a braúna (Schinopsis brasiliensis), a cássia (Senna siamea), a craibeira (Tabebuia aurea), o cumaru (Amburana cearensis), o ipê (Tabebuia impetiginosa) e o pereiro (Aspidosperma pyrifolium). De cada espécie foram retirados corpos-de-prova de 2,54 x 2,00 x 1,00 cm, com a maior dimensão na direção das fibras, em quatro posições na direção medula-casca. As amostras foram submetidas, por 14 semanas, à ação dos fungos Postia placenta e Neolentinus lepideus. A resistência natural, com exceção da algaroba e do angico (P. placenta), da craibeira (N. lepideus) e da cássia (P. placenta e N. lepideus), foi afetada pela posição na direção medula-casca, sem estar relacionada à densidade das madeiras ensaiadas. A madeira de ipê e a madeira de cerne da aroeira e braúna foram as mais resistentes aos fungos testados. As diferenças entre a resistência natural, exceto para a aroeira e braúna, não estavam associadas à concentração de extrativos solúveis em água quente.

Resistência natural de nove madeiras do semi-árido brasileiro a fungos causadores da podridão-mole

O objetivo desta pesquisa foi avaliar a resistência de nove madeiras de ocorrência no semi-árido brasileiro a fungos de podridão-mole, em condições de laboratório. As madeiras estudadas foram algaroba (Prosopis juliflora), angico (Anadenanthera colubrina var. cebil), aroeira (Myracrodruon urundeuva), braúna (Schinopsis brasiliensis), cássia (Senna siamea), craibeira (Tabebuia aurea), cumaru (Amburana cearensis), pau-d'arco (Tabebuia impetiginosa) e pereiro (Aspidosperma pyrifolium). De cada espécie foram retirados corpos-de-prova de 3,0 x 1,5 x 0,5 cm, com a maior dimensão no sentido das fibras, em quatro posições na direção medula-casca do tronco. As amostras permaneceram por 120 dias sob a ação da microflora natural existente em solo orgânico. A resistência ao apodrecimento da aroeira, braúna e cássia não foi afetada pela posição na direção medula-casca, não esteve relacionada à densidade das madeiras ensaiadas. As madeiras de aroeira e braúna (cerne), pereiro e pau-d'arco apresentaram melhor desempenho. A resistência natural não esteve associada à concentração de extrativos solúveis em água quente.

Impactos de sistemas agroflorestais e convencional sobre a qualidade do solo no semi-árido cearense

Os benefícios gerados pelos sistemas agroflorestais (SAFs) ao ambiente são comprovados nas mais diversas regiões do mundo. No entanto, na região semi-árida cearense, são poucos os trabalhos desenvolvidos englobando as vantagens atribuídas a este tipo de sistema quanto ao uso do solo. Objetivou-se avaliar o impacto de quatro SAFs e um convencional sobre a qualidade do solo, comparativamente à condição natural (caatinga nativa), após cinco anos de uso na região semi-árida cearense. Os tratamentos testados foram: agrossilvipastoril (AGP); silvipastoril (SILV); tradicional cultivado em 1998 e 1999 (TR98); tradicional cultivado em 2002 (TR02); e cultivo intensivo (CI) e duas áreas de mata nativa (MN-1 e MN-2), que foram usadas como referência de um estado de equilíbrio. O impacto dos sistemas de manejo sobre a qualidade do solo foi medido pelas seguintes variáveis: a granulometria; argila dispersa; a estabilidade e distribuição porcentual dos agregados; os teores de Ca2+, Mg2+, K+, Na+ trocáveis; a acidez potencial; o pH do solo; e o teor de carbono orgânico total (COT), considerando-se as profundidades de 0-6, 6-12, 12-20 e 20-40 cm. Os tratamentos AGP, TR98 e CI promoveram maior revolvimento do solo provocando a redução nos teores de COT. Os atributos químicos, como bases trocáveis, capacidade de troca catiônica, pH e acidez potencial nos tratamentos SILV e MN-2, variaram em função dos teores de argila, enquanto nos tratamentos TR98 e CI a redução nesses atributos foi devido ao maior revolvimento do solo. O tratamento AGP mostrou-se eficiente na ciclagem de nutrientes, entretanto, o revolvimento do solo e a concomitante redução do teor de COT, geraram também diminuição na estabilidade de agregados. Os resultados obtidos permitem recomendar o tratamento SILV para a manutenção da qualidade do solo e produção de alimentos na região do semi-árido cearense.

Produção e distribuição de biomassa de espécies arbóreas no semi-árido brasileiro

Com o objetivo de avaliar a produção e distribuição de biomassa de algumas espécies arbóreas de múltiplo uso em condições de sequeiro do submédio do São Francisco, instalou-se um experimento no Campo Experimental da Caatinga da Embrapa Semi-Árido, Município de Petrolina, PE. Foram realizadas medições de altura e diâmetro à altura do peito (DAP) de 16 árvores centrais, em três parcelas de cada espécie: Leucaena diversifolia, Caesalpinia velutina, Caesalpinia coriaria, Mimosa tenuiflora e Ateleia herbert-smithii. A biomassa foi estimada com base na árvore de altura média de cada parcela, avaliando-se, separadamente, cada componente (folhas, galhos, cascas e lenho). A biomassa nos diferentes componentes arbóreos das espécies foi distribuída na seguinte ordem: folha

Planting density and initial growth of two tree species adapted to the semi-arid region

Planting densities influence several aspects of forest formation, including management practices, timber yield, quality, and extraction, and consequently its production costs. The objective of this study was to evaluate Mimosa caesalpiinifolia and Gliricidia sepium growth as a function of planting density (400, 600, 800, 1000, and 1200 plants ha-1) and plant age. The species were evaluated every 90 days for plant height (PH), crown diameter (CD) and root collar diameter (RCD) (10 cm above the ground), with the first evaluation performed at 90 days and the last at 720 days. When plants were one year of age and beyond, evaluations were conducted also for stem diameter at breast height (DBH) (1.30 m above the ground). A randomized block design with split-plots and three replicates was adopted. Species were assigned to plots, planting densities were assigned to subplots, and evaluation ages were assigned to subsubplots. The four traits in both species had their values decreased as planting density increased, but continually increased as plant age increased. For PH and RCD there was an alternation between species superiority, with gliricidia being superior to sabiá at some ages, while the opposite occurred at other ages. As to CD the species only differed in the last measurement, gliricidia being superior. With regard to DBH, gliricidia was superior starting from the second measurement. There was an effect of the species × ages interaction for the four traits and also an effect of the densities × ages interaction for CD and DBH.

Bayesian methods for estimation, optimization and experimental design

Mathematical models often contain parameters that need to be calibrated from measured data. The emergence of efficient Markov Chain Monte Carlo (MCMC) methods has made the Bayesian approach a standard tool in quantifying the uncertainty in the parameters. With MCMC, the parameter estimation problem can be solved in a fully statistical manner, and the whole distribution of the parameters can be explored, instead of obtaining point estimates and using, e.g., Gaussian approximations. In this thesis, MCMC methods are applied to parameter estimation problems in chemical reaction engineering, population ecology, and climate modeling. Motivated by the climate model experiments, the methods are developed further to make them more suitable for problems where the model is computationally intensive. After the parameters are estimated, one can start to use the model for various tasks. Two such tasks are studied in this thesis: optimal design of experiments, where the task is to design the next measurements so that the parameter uncertainty is minimized, and model-based optimization, where a model-based quantity, such as the product yield in a chemical reaction model, is optimized. In this thesis, novel ways to perform these tasks are developed, based on the output of MCMC parameter estimation. A separate topic is dynamical state estimation, where the task is to estimate the dynamically changing model state, instead of static parameters. For example, in numerical weather prediction, an estimate of the state of the atmosphere must constantly be updated based on the recently obtained measurements. In this thesis, a novel hybrid state estimation method is developed, which combines elements from deterministic and random sampling methods.

Kernel-Based Ranking. Methods for Learning and Performance Estimation

Machine learning provides tools for automated construction of predictive models in data intensive areas of engineering and science. The family of regularized kernel methods have in the recent years become one of the mainstream approaches to machine learning, due to a number of advantages the methods share. The approach provides theoretically well-founded solutions to the problems of under- and overfitting, allows learning from structured data, and has been empirically demonstrated to yield high predictive performance on a wide range of application domains. Historically, the problems of classification and regression have gained the majority of attention in the field. In this thesis we focus on another type of learning problem, that of learning to rank. In learning to rank, the aim is from a set of past observations to learn a ranking function that can order new objects according to how well they match some underlying criterion of goodness. As an important special case of the setting, we can recover the bipartite ranking problem, corresponding to maximizing the area under the ROC curve (AUC) in binary classification. Ranking applications appear in a large variety of settings, examples encountered in this thesis include document retrieval in web search, recommender systems, information extraction and automated parsing of natural language. We consider the pairwise approach to learning to rank, where ranking models are learned by minimizing the expected probability of ranking any two randomly drawn test examples incorrectly. The development of computationally efficient kernel methods, based on this approach, has in the past proven to be challenging. Moreover, it is not clear what techniques for estimating the predictive performance of learned models are the most reliable in the ranking setting, and how the techniques can be implemented efficiently. The contributions of this thesis are as follows. First, we develop RankRLS, a computationally efficient kernel method for learning to rank, that is based on minimizing a regularized pairwise least-squares loss. In addition to training methods, we introduce a variety of algorithms for tasks such as model selection, multi-output learning, and cross-validation, based on computational shortcuts from matrix algebra. Second, we improve the fastest known training method for the linear version of the RankSVM algorithm, which is one of the most well established methods for learning to rank. Third, we study the combination of the empirical kernel map and reduced set approximation, which allows the large-scale training of kernel machines using linear solvers, and propose computationally efficient solutions to cross-validation when using the approach. Next, we explore the problem of reliable cross-validation when using AUC as a performance criterion, through an extensive simulation study. We demonstrate that the proposed leave-pair-out cross-validation approach leads to more reliable performance estimation than commonly used alternative approaches. Finally, we present a case study on applying machine learning to information extraction from biomedical literature, which combines several of the approaches considered in the thesis. The thesis is divided into two parts. Part I provides the background for the research work and summarizes the most central results, Part II consists of the five original research articles that are the main contribution of this thesis.