This study aimed to develop a machine learning-based algorithm for glaucoma diagnosis in patients with open-angle glaucoma, based on three-dimensional optical coherence tomography (OCT) data and color fundus images. In this study, 208 glaucomatous and 149 healthy eyes were enrolled, and color fundus images and volumetric OCT data from the optic disc and macular area of these eyes were captured with a spectral-domain OCT (3D OCT-2000, Topcon). Thickness and deviation maps were created with a segmentation algorithm. Transfer learning of convolutional neural network (CNN) was used with the following types of input images: (1) fundus image of optic disc in grayscale format, (2) disc retinal nerve fiber layer (RNFL) thickness map, (3) macular ganglion cell complex (GCC) thickness map, (4) disc RNFL deviation map, and (5) macular GCC deviation map. Data augmentation and dropout were performed to train the CNN. For combining the results from each CNN model, a random forest (RF) was trained to classify the disc fundus images of healthy and glaucomatous eyes using feature vector representation of each input image, removing the second fully connected layer. The area under receiver operating characteristic curve (AUC) of a 10-fold cross validation (CV) was used to evaluate the models. The 10-fold CV AUCs of the CNNs were 0.940 for color fundus images, 0.942 for RNFL thickness maps, 0.944 for macular GCC thickness maps, 0.949 for disc RNFL deviation maps, and 0.952 for macular GCC deviation maps. The RF combining the five separate CNN models improved the 10-fold CV AUC to 0.963. Therefore, the machine learning system described here can accurately differentiate between healthy and glaucomatous subjects based on their extracted images from OCT data and color fundus images. This system should help to improve the diagnostic accuracy in glaucoma.
PurposeAlthough the lamina cribrosa (LC) is the primary site of axonal damage in glaucoma, adequate methods to image and measure it are currently lacking. Here, we describe a noninvasive, in vivo method of evaluating the LC, based on swept-source optical coherence tomography (SS-OCT), and determine this method’s ability to quantify LC thickness.MethodsThis study comprised 54 eyes, including normal (n = 18), preperimetric glaucoma (PPG; n = 18), and normal tension glaucoma (NTG; n = 18) eyes. We used SS-OCT to obtain 3 x 3 mm cube scans of an area centered on the optic disc, and then synchronized reconstructed B- and en-face images from this data. We identified the LC in these B-scan images by marking the visible borders of the LC pores. We marked points on the anterior and posterior borders of the LC in 12 B-scan images in order to create a skeleton model of the LC. Finally, we used B-spline interpolation to form a 3D model of the LC, including only reliably measured scan areas. We calculated the average LC thickness (avgLCT) in this model and used Spearman's rank correlation coefficient to compare it with circumpapillary retinal nerve fiber layer thickness (cpRNFLT).ResultsWe found that the correlation coefficient of avgLCT and cpRNFLT was 0.64 (p < 0.01). The coefficient of variation for avgLCT was 5.1%. AvgLCT differed significantly in the groups (normal: 282.6 ± 20.6 μm, PPG: 261.4 ± 15.8 μm, NTG: 232.6 ± 33.3 μm). The normal, PPG and NTG groups did not significantly differ in age, sex, refractive error or intraocular pressure (IOP), although the normal and NTG groups differed significantly in cpRNFLT and Humphrey field analyzer measurements of mean deviation.ConclusionThus, our results indicate that the parameters of our newly developed method of measuring LC thickness with SS-OCT may provide useful and important data for glaucoma diagnosis and research.
PurposeTo investigate the influence of various risk factors on thinning of the lamina cribrosa (LC), as measured with swept-source optical coherence tomography (SS-OCT; Topcon).MethodsThis retrospective study comprised 150 eyes of 150 patients: 22 normal subjects, 28 preperimetric glaucoma (PPG) patients, and 100 open-angle glaucoma patients. Average LC thickness was determined in a 3 x 3 mm cube scan of the optic disc, over which a 4 x 4 grid of 16 points was superimposed (interpoint distance: 175 μm), centered on the circular Bruch’s membrane opening. The borders of the LC were defined as the visible limits of the LC pores. The correlation of LC thickness with Humphrey field analyzer-measured mean deviation (MD; SITA standard 24–2), circumpapillary retinal nerve fiber layer thickness (cpRNFLT), the vertical cup-to-disc (C/D) ratio, and tissue mean blur rate (MBR) was determined with Spearman's rank correlation coefficient. The relationship of LC thickness with age, axial length, intraocular pressure (IOP), MD, the vertical C/D ratio, central corneal thickness (CCT), and tissue MBR was determined with multiple regression analysis. Average LC thickness and the correlation between LC thickness and MD were compared in patients with the glaucomatous enlargement (GE) optic disc type and those with non-GE disc types, as classified with Nicolela’s method.ResultsWe found that average LC thickness in the 16 grid points was significantly associated with overall LC thickness (r = 0.77, P < 0.001). The measurement time for this area was 12.4 ± 2.4 minutes. Average LC thickness in this area had a correlation coefficient of 0.57 with cpRNFLT (P < 0.001) and 0.46 (P < 0.001) with MD. Average LC thickness differed significantly between the groups (normal: 268 ± 23 μm, PPG: 248 ± 13 μm, OAG: 233 ± 20 μm). Multiple regression analysis showed that MD (β = 0.29, P = 0.013), vertical C/D ratio (β = -0.25, P = 0.020) and tissue MBR (β = 0.20, P = 0.034) were independent variables significantly affecting LC thickness, but age, axial length, IOP, and CCT were not. LC thickness was significantly lower in the GE patients (233.9 ± 17.3 μm) than the non-GE patients (243.6 ± 19.5 μm, P = 0.040). The correlation coefficient between MD and LC thickness was 0.58 (P < 0.001) in the GE patients and 0.39 (P = 0.013) in the non-GE patients.ConclusionCupping formation and tissue blood flow were independently correlated to LC thinning. Glaucoma patients with the GE disc type, who predominantly have large cupping, had lower LC thickness even with similar glaucoma severity.
This study develops an objective machine-learning classification model for classifying glaucomatous optic discs and reveals the classificatory criteria to assist in clinical glaucoma management. In this study, 163 glaucoma eyes were labelled with four optic disc types by three glaucoma specialists and then randomly separated into training and test data. All the images of these eyes were captured using optical coherence tomography and laser speckle flowgraphy to quantify the ocular structure and blood-flow-related parameters. A total of 91 parameters were extracted from each eye along with the patients' background information. Machine-learning classifiers, including the neural network (NN), naïve Bayes (NB), support vector machine (SVM), and gradient boosted decision trees (GBDT), were trained to build the classification models, and a hybrid feature selection method that combines minimum redundancy maximum relevance and genetic-algorithm-based feature selection was applied to find the most valid and relevant features for NN, NB, and SVM. A comparison of the performance of the three machine-learning classification models showed that the NN had the best classification performance with a validated accuracy of 87.8% using only nine ocular parameters. These selected quantified parameters enabled the trained NN to classify glaucomatous optic discs with relatively high performance without requiring color fundus images.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.