Distancia KullBack Leibler entre distribuciones de dimensión fractal: cuantificación de la relevancia patológica de las imágenes retinianas
Loading...
Official URL
Full text at PDC
Publication date
2021
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Citation
Abstract
Se ha desarrollado un método de cuantificación del grado de patología de imagen de fondo de ojo mediante el análisis de la dimensión fractal conectada de la vasculatura retiniana. El método se basa en la cuantificación de la diferencia entre la distribución de probabilidad de esta dimensión para retinas sanas y patológicas, usando la divergencia de Kullback Leibler entre ellas. Los resultados sobre imágenes de retinopatía diabética y glaucoma muestran una alta fiabilidad respecto a la detección de patología por expertos humanos, teniendo en cuenta, además, la variabilidad propia entre retinas sanas y permitiendo cuantificar el grado de patología.
A method for quantifying the degree of retinal image pathology has been developed by analyzing the connected fractal dimension of the retinal vasculature. The method is based on the quantification of the difference between the probability distribution of this dimension for healthy and pathological retinas, using the Kullback Leibler divergence between them. The results for images of diabetic retinopathy and glaucoma show a high reliability with respect to the detection of pathology by human experts, taking into account, in addition, the own variability between healthy retinas. Moreover, the method quantifies the degree of pathology.
A method for quantifying the degree of retinal image pathology has been developed by analyzing the connected fractal dimension of the retinal vasculature. The method is based on the quantification of the difference between the probability distribution of this dimension for healthy and pathological retinas, using the Kullback Leibler divergence between them. The results for images of diabetic retinopathy and glaucoma show a high reliability with respect to the detection of pathology by human experts, taking into account, in addition, the own variability between healthy retinas. Moreover, the method quantifies the degree of pathology.