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.
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.
It is difficult to identify glaucoma in myopic eyes because the configuration of the optic disc varies; yet it is important clinically. Here, we used laser speckle flowgraphy (LSFG) to measure mean blur rate (MBR), representing optic disc microcirculation, and assessed its ability to identify glaucoma in eyes with myopic optic discs. 129 eyes (normal disc: 21 eyes; myopic disc: 108 eyes) were enrolled. The eyes were classified as normal or mildly, moderately, or severely glaucomatous with standard automated perimetry (SAP). We determined the relationship between optic nerve head (ONH) MBR, measured with LSFG, mean deviation (MD), measured with SAP, and circumpapillary retinal nerve fiber layer thickness (cpRNFLT), measured with optical coherence tomography (OCT). ONH MBR and cpRNFLT decreased significantly with the severity of glaucoma. MBR was significantly correlated with cpRNFLT and MD (r =0.65 and r =0.63, respectively). A multiple regression analysis revealed that MBR and cpRNFLT were independent factors indicating glaucoma severity. A logistic regression analysis revealed that MBR and cpRNFLT were also independent factors indicating the presence of glaucoma. In a receiver operating characteristic (ROC) analysis, MBR and cpRNFLT could both differentiate between normal and glaucomatous eyes (MBR area under the ROC curve: 0.86, with a cut-off score of 24.0 AU). These results suggest that in addition to cpRNFLT, non-invasive and objective LSFG measurements of MBR may enable the identification of glaucoma and the classification of its severity in eyes with myopic optic discs.
Early diagnosis and detection of disease progression are critical to successful therapeutic intervention in glaucoma, the leading cause of irreversible blindness worldwide. Optical coherence tomography (OCT) is a non-invasive imaging technique that allows objective quantification in vivo of key glaucomatous structural changes in the retina and the optic nerve head (ONH). Advances in OCT technology have increased the scan speed and enhanced image quality, contributing to early glaucoma diagnosis and monitoring, as well as the visualization of critically important structures deep within the ONH, such as the lamina cribrosa. OCT angiography (OCTA) is a dye-free technique for noninvasively assessing ocular microvasculature, including capillaries within each plexus serving the macula, peripapillary retina and ONH regions, as well as the deeper vessels of the choroid. This layer-specific assessment of the microvasculature has provided evidence that retinal and choroidal vascular impairments can occur during early stages of glaucoma, suggesting that OCTA-derived measurements could be used as biomarkers for enhancing detection of glaucoma and its progression, as well as to reveal novel insights about pathophysiology. Moreover, these innovations have demonstrated that damage to the macula, a critical region for the vision-related quality of life, can be observed in the early stages of glaucomatous eyes, leading to a paradigm shift in glaucoma monitoring. Other advances in software and hardware, such as artificial intelligence-based algorithms, adaptive optics, and visible-light OCT, may further benefit clinical management of glaucoma in the future. This article reviews the utility of OCT and OCTA for glaucoma diagnosis and disease progression detection, emphasizes the importance of detecting macula damage in glaucoma, and highlights the future perspective of OCT and OCTA. We conclude that the OCT and OCTA are essential glaucoma detection and monitoring tools, leading to clinical and economic benefits for patients and society.
1 November 2022 in Journal of Clinical Investigation
Normal-tension glaucoma (NTG) is a heterogeneous disease characterized by retinal ganglion cell (RGC) death leading to cupping of the optic nerve head and visual field loss at normal intraocular pressure (IOP). The pathogenesis of NTG remains unclear. Here, we describe a single nucleotide mutation in exon 2 of the methyltransferase-like 23 (METTL23) gene identified in 3 generations of a Japanese family with NTG. This mutation caused METTL23 mRNA aberrant splicing, which abolished normal protein production and altered subcellular localization. Mettl23–knock-in (Mettl23+/G and Mettl23G/G) and -knockout (Mettl23+/– and Mettl23–/–) mice developed a glaucoma phenotype without elevated IOP. METTL23 is a histone arginine methyltransferase expressed in murine and macaque RGCs. However, the novel mutation reduced METTL23 expression in RGCs of Mettl23G/G mice, which recapitulated both clinical and biological phenotypes. Moreover, our findings demonstrated that METTL23 catalyzed the dimethylation of H3R17 in the retina and was required for the transcription of pS2, an estrogen receptor α target gene that was critical for RGC homeostasis through the negative regulation of NF-κB–mediated TNF-α and IL-1β feedback. These findings suggest an etiologic role of METTL23 in NTG with tissue-specific pathology.
Dendrite pathology and synaptic loss result in neural circuit dysfunction, a common feature of neurodegenerative diseases. There is a lack of strategies that target dendritic and synaptic regeneration to promote neurorecovery. We show that daily human recombinant insulin eye drops stimulate retinal ganglion cell (RGC) dendrite and synapse regeneration during ocular hypertension, a risk factor to develop glaucoma. We demonstrate that the ribosomal protein p70S6 kinase (S6K) is essential for insulin-dependent dendritic regrowth. Furthermore, S6K phosphorylation of the stress-activated protein kinase–interacting protein 1 (SIN1), a link between the mammalian target of rapamycin complexes 1 and 2 (mTORC1/2), is required for insulin-induced dendritic regeneration. Using two-photon microscopy live retinal imaging, we show that insulin rescues single-RGC light-evoked calcium (Ca 2+ ) dynamics. We further demonstrate that insulin enhances neuronal survival and retina-brain connectivity leading to improved optomotor reflex–elicited behaviors. Our data support that insulin is a compelling pro-regenerative strategy with potential clinical implications for the treatment and management of glaucoma.