Análisis de la capa de células ganglionares con deep learning en el diagnóstico de glaucoma / Ganglion cell layer analysis with deep learning in glaucoma diagnosis

Objetivo Determinar y comparar la precisión diagnóstica en glaucoma de dos modelos de aprendizaje profundo, usando imágenes en infrarrojo del nervio óptico, del fondo de ojo y de la capa de células ganglionares (CCG). Métodos Hemos seleccionado una muestra de pacientes normales y con glaucoma. Se recogieron tres imágenes en infrarrojo con un tomógrafo de coherencia óptica de tipo spectral-domain (SD-OCT). La primera corresponde a la imagen de barrido confocal del fondo de ojo, la segunda es un recorte de la primera centrada en el nervio óptico, y la tercera fue la imagen del corte SD-OCT de la CCG. Nuestros modelos de aprendizaje profundo se desarrollaron en la plataforma MATLAB con las redes neuronales preentrenadas ResNet50 y VGG19. Resultados Se recogieron 498 ojos de 298 pacientes. De los 498 ojos, 312 son glaucomatosos y 186 son normales. En la prueba, la precisión de los modelos fue de 96% (ResNet50) y 96% (VGG19) para las imágenes de la CCG, de 90% (ResNet50) y 90% (VGG19) para las imágenes del nervio óptico y de 82% (ResNet50) y 84% (VGG19) para las de fondo de ojo. El área ROC en la prueba fue de 0,96 (ResNet50) y 0,97 (VGG19) para las imágenes de la CCG, de 0,87 (ResNet50) y 0,88 (VGG19) para las imágenes del nervio óptico, y de 0,79 (ResNet50) y 0,81 (VGG19) para las imágenes de fondo de ojo. Conclusiones Los dos modelos de aprendizaje profundo analizados, aplicados sobre las imágenes de la CCG, ofrecen una alta precisión diagnóstica, sensibilidad y especificidad en el diagnóstico de glaucoma (AU)

Objective To determine and compare the diagnostic precision in glaucoma of two deep learning models using infrared images of the optic nerve, eye fundus, and the ganglion cell layer (GCL). Methods We have selected a sample of normal and glaucoma patients. Three infrared images were registered with a spectral-domain optical coherence tomography (SD-OCT). The first corresponds to the confocal scan image of the fundus, the second is a cut-out of the first centered on the optic nerve, and the third was the SD-OCT image of the GCL. Our deep learning models are developed on the MatLab platform with the ResNet50 and VGG19 pre-trained neural networks. Results 498 eyes of 298 patients were collected. Of the 498 eyes, 312 are glaucoma and 186 are normal. In the test, the precision of the models was 96% (ResNet50) and 96% (VGG19) for the GCL images, 90% (ResNet50) and 90% (VGG19) for the optic nerve images and 82% (ResNet50) and 84% (VGG19) for the fundus images. The ROC area in the test was 0.96 (ResNet50) and 0.97 (VGG19) for the GCL images, 0.87 (ResNet50) and 0.88 (VGG19) for the optic nerve images, and 0.79 (ResNet50) and 0.81 (VGG19) for the fundus images. Conclusions Both deep learning models, applied to the GCL images, achieve high diagnostic precision, sensitivity and specificity in the diagnosis of glaucoma (AU)

Ganglion cell layer analysis with deep learning in glaucoma diagnosis. / Análisis de la capa de células ganglionares con deep learning en el diagnóstico de glaucoma.

OBJECTIVE: To determine and compare the diagnostic precision in glaucoma of two deep learning models using infrared images of the optic nerve, eye fundus, and the ganglion cell layer (GCL). METHODS: We have selected a sample of normal and glaucoma patients. Three infrared images were registered with a spectral-domain optical coherence tomography (SD-OCT). The first corresponds to the confocal scan image of the fundus, the second is a cut-out of the first centered on the optic nerve, and the third was the SD-OCT image of the GCL. Our deep learning models are developed on the MatLab platform with the ResNet50 and VGG19 pre-trained neural networks. RESULTS: 498 eyes of 298 patients were collected. Of the 498 eyes, 312 are glaucoma and 186 are normal. In the test, the precision of the models was 96% (ResNet50) and 96% (VGG19) for the GCL images, 90% (ResNet50) and 90% (VGG19) for the optic nerve images and 82% (ResNet50) and 84% (VGG19) for the fundus images. The ROC area in the test was 0.96 (ResNet50) and 0.97 (VGG19) for the GCL images, 0.87 (ResNet50) and 0.88 (VGG19) for the optic nerve images, and 0.79 (ResNet50) and 0.81 (VGG19) for the fundus images. CONCLUSIONS: Both deep learning models, applied to the GCL images, achieve high diagnostic precision, sensitivity and specificity in the diagnosis of glaucoma.

Estimating the Amount of Hemoglobin in the Neuroretinal Rim Using Color Images and OCT.

PURPOSE: To calculate the amount of hemoglobin (Hb) in the optic nerve head (ONH), using superimposed color fundus images with disc, rim and cup boundaries obtained by OCT-Cirrus. MATERIAL AND METHODS: We examined 100 healthy and 121 glaucomatous eyes using Oculus-Spark perimetry, Cirrus-OCT and Visucam (Zeiss) ONH color images. The Laguna ONhE program was then used to calculate the amount of Hb in the cup and six sectors of the rim. Receiver operating characteristic (ROC) analysis was performed and correlations between parameters were calculated. RESULTS: In suspected and confirmed glaucoma, Hb was significantly lower than controls in all rim sectors, especially the inferior and superonasal (p < 0.0001). Mean deviation (MD) of visual field regions showed greater correlation with the amount of Hb in the superior and inferior sectors of the rim than with rim area (p = 0.02) or nerve fiber layer thickness (p < 0.0001). On ROC analysis, the best diagnostic indicators were OCT rim area, vertical cup/disc ratio (C/D) and Glaucoma Discriminant Function (GDF) of Laguna ONhE, without significant differences. CONCLUSIONS: The amount of Hb in the ONH seems to have an important relationship with glaucomatous visual field sensitivity. The remaining rim has insufficient perfusion in many cases of glaucoma.

Estimation of the relative amount of hemoglobin in the cup and neuroretinal rim using stereoscopic color fundus images.

PURPOSE: To calculate the relative amount of hemoglobin (Hb) in sectors of the optic nerve head (ONH) from stereoscopic color fundus images using the Laguna ONhE method and compare the results with the visual field evaluation and optical coherence tomography (OCT). METHODS: Healthy eyes (n = 87) and glaucoma eyes (n = 71) underwent reliable Oculus Spark perimetry and Cirrus OCT. Optical nerve head color images were acquired with a nonmydriatic stereoscopic Wx Kowa fundus camera. Laguna ONhE program was applied to these images to calculate the relative Hb amount in the cup and six sectors of the rim. Receiver operating characteristic (ROC) analysis and correlations between parameters were calculated. RESULTS: We did not observe any variations in the relative amount of Hb in relation to age in healthy subjects (R(2) = 0.033, P > 0.05). Maximum ROC area confidence intervals were observed for a combination between perimetric indices and the Laguna ONhE Glaucoma discriminant function (0.970-0.899) followed by rim area (0.960-0.883), and mean deviation (MD; 0.944-0.857). In glaucoma cases, relative Hb amount presented significant reduction in all rim sectors, especially 231° to 270° and 81° to 120° (P < 0.001), except in the temporal 311° to 40° (P = 0.11). Perimetry mean sensitivity by sectors was better correlated with respective Hb levels than with rim areas or the corresponding nerve fiber thickness, especially the superior and inferior sectors (P < 0.05). CONCLUSIONS: Visual field sensitivity was better correlated with Hb levels than with rim sector areas or the corresponding nerve fiber thickness. In many cases the remaining rim show low perfusion, especially in the superior and inferior sectors.