Neural networks and statistical analysis for classification of corneal videokeratography maps based on Zernike coefficients: a quantitative comparison.

PURPOSE: The main goal of this study was to develop and compare two different techniques for classification of specific types of corneal shapes when Zernike coefficients are used as inputs. A feed-forward artificial Neural Network (NN) and discriminant analysis (DA) techniques were used. METHODS: The inputs both for the NN and DA were the first 15 standard Zernike coefficients for 80 previously classified corneal elevation data files from an Eyesys System 2000 Videokeratograph (VK), installed at the Departamento de Oftalmologia of the Escola Paulista de Medicina, São Paulo. The NN had 5 output neurons which were associated with 5 typical corneal shapes: keratoconus, with-the-rule astigmatism, against-the-rule astigmatism, "regular" or "normal" shape and post-PRK. RESULTS: The NN and DA responses were statistically analyzed in terms of precision ([true positive+true negative]/total number of cases). Mean overall results for all cases for the NN and DA techniques were, respectively, 94% and 84.8%. CONCLUSION: Although we used a relatively small database, results obtained in the present study indicate that Zernike polynomials as descriptors of corneal shape may be a reliable parameter as input data for diagnostic automation of VK maps, using either NN or DA.

Neuromorphometric characterization with shape functionals.

This work presents a procedure to extract morphological information from neuronal cells based on the variation of shape functionals as the cell geometry undergoes a dilation through a wide interval of spatial scales. The targeted shapes are alpha and beta cat retinal ganglion cells, which are characterized by different ranges of dendritic field diameter. Image functionals are expected to act as descriptors of the shape, gathering relevant geometric and topological features of the complex cell form. We present a comparative study of classification performance of additive shape descriptors, namely, Minkowski functionals, and the nonadditive multiscale fractal. We found that the proposed measures perform efficiently the task of identifying the two main classes alpha and beta based solely on scale invariant information, while also providing intraclass morphological assessment.

Neural networks and statistical analysis for classification of corneal videokeratography maps based on Zernike coefficients: a quantitative comparison / Redes neurais e análise estatística para classificação de mapas topográficos da córnea baseados em coeficientes de Zernike: uma comparação quantitativa

PURPOSE: The main goal of this study was to develop and compare two different techniques for classification of specific types of corneal shapes when Zernike coefficients are used as inputs. A feed-forward artificial Neural Network (NN) and discriminant analysis (DA) techniques were used. METHODS: The inputs both for the NN and DA were the first 15 standard Zernike coefficients for 80 previously classified corneal elevation data files from an Eyesys System 2000 Videokeratograph (VK), installed at the Departamento de Oftalmologia of the Escola Paulista de Medicina, São Paulo. The NN had 5 output neurons which were associated with 5 typical corneal shapes: keratoconus, with-the-rule astigmatism, against-the-rule astigmatism, "regular" or "normal" shape and post-PRK. RESULTS: The NN and DA responses were statistically analyzed in terms of precision ([true positive+true negative]/total number of cases). Mean overall results for all cases for the NN and DA techniques were, respectively, 94 percent and 84.8 percent. CONCLUSION: Although we used a relatively small database, results obtained in the present study indicate that Zernike polynomials as descriptors of corneal shape may be a reliable parameter as input data for diagnostic automation of VK maps, using either NN or DA.

OBJETIVOS: Nosso principal objetivo neste trabalho foi de desenvolver e comparar duas técnicas diferentes para classificação de superfícies corneanas. Uma rede neural artificial alimentada adiante e análise descriminante foram as técnicas de classificação comparadas neste trabalho. MÉTODOS: As entradas para ambos os métodos de classificação foram os primeiros 15 coeficientes de Zernike para 80 córneas mensuradas anteriormente em um topógrafo Eyesys instalado no Departamento de Oftalmologia da Escola Paulista de Medicina - UNIFESP. A rede neural tem 5 saídas que foram associados aos cinco casos típicos contidos na base de dados: ceratocone, astigmatismo a favor da regra, astigmatismo contra a regra, formato "regular" ou "normal" e pós-PRK. RESULTADOS: Os resultados de ambos os métodos foram estatisticamente analisados em termos de precisão. Os resultados gerais para ambos os métodos de redes neurais e análise discriminante foram 94 por cento e 84,8 por cento, respectivamente. CONCLUSÃO: Embora tenha-se utilizado uma base de dados relativamente pequena, os resultados obtidos aqui indicam que os polinômios de Zernike podem ser um parâmetro de entrada para classificação de diferentes formatos de córnea, tanto para uso com redes neurais ou análise discriminante.