Diagnostic accuracy of optic nerve head and macula OCT parameters for detecting glaucoma in eyes with and without high axial myopia.

PURPOSE: To characterize structural differences and assess the diagnostic accuracy of optic nerve head (ONH) and macula optical coherence tomography (OCT) parameters to detect glaucoma in eyes with and without high axial myopia. DESIGN: Cross-sectional study METHODS: 368 glaucoma and 411 healthy eyes with no axial myopia, 393 glaucoma and 271 healthy eyes with mild axial myopia and 124 glaucoma and 85 healthy eyes with high axial myopia were included. Global and sectoral peripapillary retinal nerve fiber layer thickness (pRNFLT), Bruch's membrane opening minimum rim width (BMO-MRW), ganglion cell inner plexiform layer thickness (GCIPLT), and macula RNFLT (mRNFLT) were compared and the diagnostic accuracy for glaucoma detection was evaluated using adjusted area under the receiver operating characteristic curve (AUC). RESULTS: Diagnostic accuracy for ONH and macula parameters to detect glaucoma was generally high and differed by myopia group. For ONH parameters the diagnostic accuracy was highest for global (AUC=0.95) and inferotemporal (AUC=0.91) pRNFLT for high myopes and global BMO-MRW for non-myopes (AUC=1.0) and mild myopes (AUC=0.97). For macula parameters, the diagnostic accuracy was higher in high myopes with 6 of the 11 GCIPLT global/sectors having adjusted AUCs > 0.90 compared to non-high myopes with no AUCs > 0.90. In all myopia groups, mRNFLT had lower AUCs than GCIPLT. CONCLUSIONS: The diagnostic accuracy for pRNFL and GCIPL was high for high axial myopic eyes and shows promise for glaucoma detection in high myopes. Further analysis is needed to determine whether the high diagnostic accuracy can be confirmed in other populations.

Proactive Decision Support for Glaucoma Treatment: Predicting Surgical Interventions with Clinically Available Data.

A longitudinal ophthalmic dataset was used to investigate multi-modal machine learning (ML) models incorporating patient demographics and history, clinical measurements, optical coherence tomography (OCT), and visual field (VF) testing in predicting glaucoma surgical interventions. The cohort included 369 patients who underwent glaucoma surgery and 592 patients who did not undergo surgery. The data types used for prediction included patient demographics, history of systemic conditions, medication history, ophthalmic measurements, 24-2 VF results, and thickness measurements from OCT imaging. The ML models were trained to predict surgical interventions and evaluated on independent data collected at a separate study site. The models were evaluated based on their ability to predict surgeries at varying lengths of time prior to surgical intervention. The highest performing predictions achieved an AUC of 0.93, 0.92, and 0.93 in predicting surgical intervention at 1 year, 2 years, and 3 years, respectively. The models were also able to achieve high sensitivity (0.89, 0.77, 0.86 at 1, 2, and 3 years, respectively) and specificity (0.85, 0.90, and 0.91 at 1, 2, and 3 years, respectively) at an 0.80 level of precision. The multi-modal models trained on a combination of data types predicted surgical interventions with high accuracy up to three years prior to surgery and could provide an important tool to predict the need for glaucoma intervention.

Deep Learning Identifies High-Quality Fundus Photographs and Increases Accuracy in Automated Primary Open Angle Glaucoma Detection.

Purpose: To develop and evaluate a deep learning (DL) model to assess fundus photograph quality, and quantitatively measure its impact on automated POAG detection in independent study populations. Methods: Image quality ground truth was determined by manual review of 2815 fundus photographs of healthy and POAG eyes from the Diagnostic Innovations in Glaucoma Study and African Descent and Glaucoma Evaluation Study (DIGS/ADAGES), as well as 11,350 from the Ocular Hypertension Treatment Study (OHTS). Human experts assessed a photograph as high quality if of sufficient quality to determine POAG status and poor quality if not. A DL quality model was trained on photographs from DIGS/ADAGES and tested on OHTS. The effect of DL quality assessment on DL POAG detection was measured using area under the receiver operating characteristic (AUROC). Results: The DL quality model yielded an AUROC of 0.97 for differentiating between high- and low-quality photographs; qualitative human review affirmed high model performance. Diagnostic accuracy of the DL POAG model was significantly greater (P < 0.001) in good (AUROC, 0.87; 95% CI, 0.80-0.92) compared with poor quality photographs (AUROC, 0.77; 95% CI, 0.67-0.88). Conclusions: The DL quality model was able to accurately assess fundus photograph quality. Using automated quality assessment to filter out low-quality photographs increased the accuracy of a DL POAG detection model. Translational Relevance: Incorporating DL quality assessment into automated review of fundus photographs can help to decrease the burden of manual review and improve accuracy for automated DL POAG detection.

Central visual field damage in glaucoma eyes with choroidal microvasculature dropout with and without high axial myopia.

PURPOSE: To characterise the relationship between a deep-layer microvasculature dropout (MvD) and central visual field (VF) damage in primary open-angle glaucoma (POAG) patients with and without high axial myopia. DESIGN: Cross-sectional study. METHODS: Seventy-one eyes (49 patients) with high axial myopia and POAG and 125 non-highly myopic POAG eyes (97 patients) were enrolled. Presence, area and angular circumference of juxtapapillary MvD were evaluated on optical coherence tomography angiography B-scans and en-face choroidal images. RESULTS: Juxtapapillary MvD was detected more often in the highly myopic POAG eyes (43 eyes, 86%) than in the non-highly myopic eyes (73 eyes, 61.9%; p=0.002). In eyes with MvD, MvD area and angular circumference (95% CI) were significantly larger in the highly myopic eyes compared with the non-highly myopic eyes (area: (0.69 (0.40, 0.98) mm2 vs 0.31 (0.19, 0.42) mm2, p=0.011) and (angular circumference: 84.3 (62.9, 105.8) vs 74.5 (58.3, 90.9) degrees, p<0.001), respectively. 24-2 VF mean deviation (MD) was significantly worse in eyes with MvD compared with eyes without MvD in both groups (p<0.001). After adjusting for 24-2 MD VF, central VF defects were more frequently found in eyes with MvD compared with eyes without MvD (82.7% vs 60.9%, p<0.001). In multivariable analysis, higher intraocular pressure, worse 24-2 VF MD, longer axial length and greater MvD area and angular circumference were associated with worse 10-2 VF MD. CONCLUSIONS: MvD was more prevalent and larger in POAG eyes with high myopia than in non-highly myopic POAG eyes. In both groups, eyes with MvD showed worse glaucoma severity and more central VF defects.

Multimodal Deep Learning Classifier for Primary Open Angle Glaucoma Diagnosis Using Wide-Field Optic Nerve Head Cube Scans in Eyes With and Without High Myopia.

PRCIS: An optical coherence tomography (OCT)-based multimodal deep learning (DL) classification model, including texture information, is introduced that outperforms single-modal models and multimodal models without texture information for glaucoma diagnosis in eyes with and without high myopia. BACKGROUND/AIMS: To evaluate the diagnostic accuracy of a multimodal DL classifier using wide OCT optic nerve head cube scans in eyes with and without axial high myopia. MATERIALS AND METHODS: Three hundred seventy-one primary open angle glaucoma (POAG) eyes and 86 healthy eyes, all without axial high myopia [axial length (AL) ≤ 26 mm] and 92 POAG eyes and 44 healthy eyes, all with axial high myopia (AL > 26 mm) were included. The multimodal DL classifier combined features of 3 individual VGG-16 models: (1) texture-based en face image, (2) retinal nerve fiber layer (RNFL) thickness map image, and (3) confocal scanning laser ophthalmoscope (cSLO) image. Age, AL, and disc area adjusted area under the receiver operating curves were used to compare model accuracy. RESULTS: Adjusted area under the receiver operating curve for the multimodal DL model was 0.91 (95% CI = 0.87, 0.95). This value was significantly higher than the values of individual models [0.83 (0.79, 0.86) for texture-based en face image; 0.84 (0.81, 0.87) for RNFL thickness map; and 0.68 (0.61, 0.74) for cSLO image; all P ≤ 0.05]. Using only highly myopic eyes, the multimodal DL model showed significantly higher diagnostic accuracy [0.89 (0.86, 0.92)] compared with texture en face image [0.83 (0.78, 0.85)], RNFL [0.85 (0.81, 0.86)] and cSLO image models [0.69 (0.63, 0.76)] (all P ≤ 0.05). CONCLUSIONS: Combining OCT-based RNFL thickness maps with texture-based en face images showed a better ability to discriminate between healthy and POAG than thickness maps alone, particularly in high axial myopic eyes.

Detecting Glaucoma from Fundus Photographs Using Deep Learning without Convolutions: Transformer for Improved Generalization.

Purpose: To compare the diagnostic accuracy and explainability of a Vision Transformer deep learning technique, Data-efficient image Transformer (DeiT), and ResNet-50, trained on fundus photographs from the Ocular Hypertension Treatment Study (OHTS) to detect primary open-angle glaucoma (POAG) and identify the salient areas of the photographs most important for each model's decision-making process. Design: Evaluation of a diagnostic technology. Subjects Participants and Controls: Overall 66 715 photographs from 1636 OHTS participants and an additional 5 external datasets of 16 137 photographs of healthy and glaucoma eyes. Methods: Data-efficient image Transformer models were trained to detect 5 ground-truth OHTS POAG classifications: OHTS end point committee POAG determinations because of disc changes (model 1), visual field (VF) changes (model 2), or either disc or VF changes (model 3) and Reading Center determinations based on disc (model 4) and VFs (model 5). The best-performing DeiT models were compared with ResNet-50 models on OHTS and 5 external datasets. Main Outcome Measures: Diagnostic performance was compared using areas under the receiver operating characteristic curve (AUROC) and sensitivities at fixed specificities. The explainability of the DeiT and ResNet-50 models was compared by evaluating the attention maps derived directly from DeiT to 3 gradient-weighted class activation map strategies. Results: Compared with our best-performing ResNet-50 models, the DeiT models demonstrated similar performance on the OHTS test sets for all 5 ground-truth POAG labels; AUROC ranged from 0.82 (model 5) to 0.91 (model 1). Data-efficient image Transformer AUROC was consistently higher than ResNet-50 on the 5 external datasets. For example, AUROC for the main OHTS end point (model 3) was between 0.08 and 0.20 higher in the DeiT than ResNet-50 models. The saliency maps from the DeiT highlight localized areas of the neuroretinal rim, suggesting important rim features for classification. The same maps in the ResNet-50 models show a more diffuse, generalized distribution around the optic disc. Conclusions: Vision Transformers have the potential to improve generalizability and explainability in deep learning models, detecting eye disease and possibly other medical conditions that rely on imaging for clinical diagnosis and management.

Macula structural and vascular differences in glaucoma eyes with and without high axial myopia.

AIMS: To identify clinically relevant parameters for identifying glaucoma in highly myopic eyes, an investigation was conducted of the relationship between the thickness of various retinal layers and the superficial vessel density (sVD) of the macula with axial length (AL) and visual field mean deviation (VFMD). METHODS: 270 glaucoma patients (438 eyes) participating in the Diagnostic Innovations in Glaucoma cross-sectional study representing three axial myopia groups (non-myopia: n=163 eyes; mild myopia: n=218 eyes; high myopia (AL>26 mm): n=57 eyes) who completed macular optical coherence tomography (OCT) and OCT-angiography imaging were included. Associations of AL and VFMD with the thickness of the ganglion cell inner plexiform layer (GCIPL), macular retinal nerve fibre layer (mRNFL), ganglion cell complex (GCC), macular choroidal thickness (mCT) and sVD were evaluated. RESULTS: Thinner Global GCIPL and GCC were significantly associated with worse VFMD (R2=34.5% and R2=32.9%; respectively p<0.001), but not with AL (all p>0.1). Thicker mRNFL showed a weak association with increasing AL (R2=2.4%; p=0.005) and a positive association with VFMD (global R2=19.2%; p<0.001). Lower sVD was weakly associated with increasing AL (R2=1.8%; p=0.028) and more strongly associated with more severe glaucoma VFMD (R2=29.6%; p<0.001). Thinner mCT was associated with increasing AL (R2=15.5% p<0.001) and not associated with VFMD (p=0.194). mRNFL was thickest while mCT was thinnest in all sectors of high myopic eyes. CONCLUSIONS: As thinner GCIPL and GCC were associated with increasing severity of glaucoma but were not significantly associated with AL, they may be useful for monitoring glaucoma in highly myopic eyes.

Comparison of Optic Disc Ovality Index and Rotation Angle Measurements in Myopic Eyes Using Photography and OCT Based Techniques.

Purpose: To compare optic nerve head (ONH) ovality index and rotation angle measurements based on semi-automated delineation of the clinical ONH margin derived from photographs and automated BMO configuration derived from optical coherence tomography (OCT) images in healthy and glaucomatous eyes with high-, mild- and no axial myopia. Methods: One hundred seventy-five healthy and glaucomatous eyes of 146 study participants enrolled in the Diagnostic Innovations in Glaucoma Study (DIGS) with optic disc photographs and Spectralis OCT ONH scans acquired on the same day were stratified by level of axial myopia (non-myopic [n = 56, axial length (AL) <24 mm], mild-myopic [n = 58, AL 24-26 mm] and high-myopic [n = 32, AL >26 mm]. The clinical disc margin of each photograph was manually annotated, and semi-automated measurements were recorded of the ovality index and rotation angle based on a best-fit ellipse generated using ImageJ software. These semi-automated photograph-based measurements were compared to ovality index and rotation angle generated from custom automated BMO-based analysis using segmented OCT ONH volumes. R 2 values from linear mixed effects models were used to describe the associations between semi-automated, photograph-based and automated OCT-based measurements. Results: Average (95% CI) axial length was 23.3 (23.0, 23.3) mm, 24.8 (24.7, 25.0) mm and 26.8 (26.6, 27.0) mm in non-myopic, mild-myopic and high-myopic eyes, respectively (ANOVA, p ≤ 0.001 for all). The R 2 association (95% CI) between semi-automated photograph-based and automated OCT-based assessment of ONH OI for all eyes was [0.26 (0.16, 0.36); p < 0.001]. This association was weakest in non-myopic eyes [0.09 (0.01, 0.26); p = 0.02], followed by mild-myopic eyes [0.13 (0.02, 0.29); p = 0.004] and strongest in high-myopic eyes [0.40 (0.19, 0.60); p < 0.001]. No significant associations were found between photography- and OCT-based assessment of rotation angle with R 2 values ranging from 0.00 (0.00, 0.08) in non-myopic eyes to 0.03 (0.00, 0.21) in high-myopic eyes (all associations p ≥ 0.33). Conclusions: Agreement between photograph-based and automated OCT-based ONH morphology measurements is limited, suggesting that these methods cannot be used interchangeably for characterizing myopic changes in the ONH.

Diagnostic Accuracy of Macular Thickness Map and Texture En Face Images for Detecting Glaucoma in Eyes With Axial High Myopia.

PURPOSE: To evaluate the diagnostic accuracy of a novel optical coherence tomography texture-based en face image analysis (SALSA-Texture) that requires segmentation of only 1 retinal layer for glaucoma detection in eyes with axial high myopia, and to compare SALSA-Texture with standard macular ganglion cell-inner plexiform layer (GCIPL) thickness, macular retinal nerve fiber layer (mRNFL) thickness, and ganglion cell complex (GCC) thickness maps. DESIGN: Comparison of diagnostic approaches. METHODS: Cross-sectional data were collected from 92 eyes with primary open-angle glaucoma (POAG) and 44 healthy control eyes with axial high myopia (axial length >26 mm). Optical coherence tomography texture en face images, developed using SALSA-Texture to model the spatial arrangement patterns of the pixel intensities in a region, were generated from 70-µm slabs just below the vitreal border of the inner limiting membrane. Areas under the receiver operating characteristic curves (AUROCs) and areas under the precision recall curves (AUPRCs) adjusted for both eyes, axial length, age, disc area, and image quality were used to compare different approaches. RESULTS: The best parameter-adjusted AUROCs (95% confidence intervals) for differentiating between healthy and glaucoma high myopic eyes were 0.92 (0.88-0.94) for texture en face images, 0.88 (0.86-0.91) for macular RNFL thickness, 0.87 (0.83-0.89) for macula GCIPL thickness, and 0.87 (0.84-0.89) for GCC thickness. A subset analysis of highly advanced myopic eyes (axial length ≥27 mm; 38 glaucomatous eyes and 22 healthy eyes) showed the best AUROC was 0.92 (0.89-0.94) for texture en face images compared with 0.86 (0.84-0.88) for macular GCIPL, 0.86 (0.84-0.88) for GCC, and 0.84 (0.81-0.87) for RNFL thickness (P ≤ .02 compared with texture for all comparisons). CONCLUSION: The current results suggest that our novel en face texture-based analysis method can improve on most investigated macular tissue thickness measurements for discriminating between highly myopic glaucomatous and highly myopic healthy eyes. While further investigation is needed, texture en face images show promise for improving the detection of glaucoma in eyes with high myopia where traditional retinal layer segmentation often is challenging.

Detecting Glaucoma in the Ocular Hypertension Study Using Deep Learning.

Importance: Automated deep learning (DL) analyses of fundus photographs potentially can reduce the cost and improve the efficiency of reading center assessment of end points in clinical trials. Objective: To investigate the diagnostic accuracy of DL algorithms trained on fundus photographs from the Ocular Hypertension Treatment Study (OHTS) to detect primary open-angle glaucoma (POAG). Design, Setting, and Participants: In this diagnostic study, 1636 OHTS participants from 22 sites with a mean (range) follow-up of 10.7 (0-14.3) years. A total of 66 715 photographs from 3272 eyes were used to train and test a ResNet-50 model to detect the OHTS Endpoint Committee POAG determination based on optic disc (287 eyes, 3502 photographs) and/or visual field (198 eyes, 2300 visual fields) changes. Three independent test sets were used to evaluate the generalizability of the model. Main Outcomes and Measures: Areas under the receiver operating characteristic curve (AUROC) and sensitivities at fixed specificities were calculated to compare model performance. Evaluation of false-positive rates was used to determine whether the DL model detected POAG before the OHTS Endpoint Committee POAG determination. Results: A total of 1147 participants were included in the training set (661 [57.6%] female; mean age, 57.2 years; 95% CI, 56.6-57.8), 167 in the validation set (97 [58.1%] female; mean age, 57.1 years; 95% CI, 55.6-58.7), and 322 in the test set (173 [53.7%] female; mean age, 57.2 years; 95% CI, 56.1-58.2). The DL model achieved an AUROC of 0.88 (95% CI, 0.82-0.92) for the OHTS Endpoint Committee determination of optic disc or VF changes. For the OHTS end points based on optic disc changes or visual field changes, AUROCs were 0.91 (95% CI, 0.88-0.94) and 0.86 (95% CI, 0.76-0.93), respectively. False-positive rates (at 90% specificity) were higher in photographs of eyes that later developed POAG by disc or visual field (27.5% [56 of 204]) compared with eyes that did not develop POAG (11.4% [50 of 440]) during follow-up. The diagnostic accuracy of the DL model developed on the optic disc end point applied to 3 independent data sets was lower, with AUROCs ranging from 0.74 (95% CI, 0.70-0.77) to 0.79 (95% CI, 0.78-0.81). Conclusions and Relevance: The model's high diagnostic accuracy using OHTS photographs suggests that DL has the potential to standardize and automate POAG determination for clinical trials and management. In addition, the higher false-positive rate in early photographs of eyes that later developed POAG suggests that DL models detected POAG in some eyes earlier than the OHTS Endpoint Committee, reflecting the OHTS design that emphasized a high specificity for POAG determination by requiring a clinically significant change from baseline.

Deep Learning Image Analysis of Optical Coherence Tomography Angiography Measured Vessel Density Improves Classification of Healthy and Glaucoma Eyes.

PURPOSE: To compare convolutional neural network (CNN) analysis of en face vessel density images to gradient boosting classifier (GBC) analysis of instrument-provided, feature-based optical coherence tomography angiography (OCTA) vessel density measurements and OCT retinal nerve fiber layer (RNFL) thickness measurements for classifying healthy and glaucomatous eyes. DESIGN: Comparison of diagnostic approaches. METHODS: A total of 130 eyes of 80 healthy individuals and 275 eyes of 185 glaucoma patients with optic nerve head (ONH) OCTA and OCT imaging were included. Classification performance of a VGG16 CNN trained and tested on entire en face 4.5 × 4.5-mm radial peripapillary capillary OCTA ONH images was compared to the performance of separate GBC models trained and tested on standard OCTA and OCT measurements. Five-fold cross-validation was used to test predictions for CNNs and GBCs. Areas under the precision recall curves (AUPRC) were calculated to control for training/test set size imbalance and were compared. RESULTS: Adjusted AUPRCs for GBC models were 0.89 (95% CI = 0.82, 0.92) for whole image vessel density GBC, 0.89 (0.83, 0.92) for whole image capillary density GBC, 0.91 (0.88, 0.93) for combined whole image vessel and whole image capillary density GBC, and 0.93 (0.91, 095) for RNFL thickness GBC. The adjusted AUPRC using CNN analysis of en face vessel density images was 0.97 (0.95, 0.99) resulting in significantly improved classification compared to GBC OCTA-based results and GBC OCT-based results (P ≤ 0.01 for all comparisons). CONCLUSION: Deep learning en face image analysis improves on feature-based GBC models for classifying healthy and glaucoma eyes.

Bruch Membrane Opening Detection Accuracy in Healthy Eyes and Eyes With Glaucoma With and Without Axial High Myopia in an American and Korean Cohort.

PURPOSE: To determine the predictors of Bruch membrane opening (BMO) location accuracy and the visibility of the BMO location in glaucoma and healthy individuals with and without axial high myopia. DESIGN: Cross-sectional study. METHODS: Healthy eyes and eyes with glaucoma from an American study and a Korean clinic population were classified into 2 groups: those with no axial high myopia (axial length [AL] <26 mm) and those with axial high myopia (AL ≥26 mm). The accuracy of the automated BMO location on optic nerve head Spectralis optical coherence tomography radial scans was assessed by expert reviewers. RESULTS: Four hundred thirty-eight non-highly myopic eyes (263 subjects) and 113 highly myopic eyes (81 subjects) were included. In healthy eyes with and without axial high myopia, 9.1% and 1.7% had indiscernible BMOs while 54.5% and 87.6% were accurately segmented, respectively. More than a third (36.4%) and 10.7% of eyes with indiscernible BMOs were manually correctable (respectively, P = .017). In eyes with glaucoma with and without high myopia, 15.0% and 3.2% had indiscernible BMOs, 55.0% and 38.2% were manually corrected, and 30.0% and 58.7% were accurately segmented without the need for manual correction (respectively, P = .005). Having axial high myopia, a larger AL, a larger BMO tilt angle, a lower BMO ovality index (more oval), and a glaucoma diagnosis were significant predictors of BMO location inaccuracy in multivariable logistic regression analysis. CONCLUSIONS: As BMO location inaccuracy was 2.4 times more likely in eyes with high axial myopia regardless of diagnosis, optical coherence tomography images of high myopes should be reviewed carefully, and when possible, BMO location should be corrected before using optic nerve head scan results for the clinical management of glaucoma.

Individualized Glaucoma Change Detection Using Deep Learning Auto Encoder-Based Regions of Interest.

Purpose: To compare change over time in eye-specific optical coherence tomography (OCT) retinal nerve fiber layer (RNFL)-based region-of-interest (ROI) maps developed using unsupervised deep-learning auto-encoders (DL-AE) to circumpapillary RNFL (cpRNFL) thickness for the detection of glaucomatous progression. Methods: Forty-four progressing glaucoma eyes (by stereophotograph assessment), 189 nonprogressing glaucoma eyes (by stereophotograph assessment), and 109 healthy eyes were followed for ≥3 years with ≥4 visits using OCT. The San Diego Automated Layer Segmentation Algorithm was used to automatically segment the RNFL layer from raw three-dimensional OCT images. For each longitudinal series, DL-AEs were used to generate individualized eye-based ROI maps by identifying RNFL regions of likely progression and no change. Sensitivities and specificities for detecting change over time and rates of change over time were compared for the DL-AE ROI and global cpRNFL thickness measurements derived from a 2.22-mm to 3.45-mm annulus centered on the optic disc. Results: The sensitivity for detecting change in progressing eyes was greater for DL-AE ROIs than for global cpRNFL annulus thicknesses (0.90 and 0.63, respectively). The specificity for detecting not likely progression in nonprogressing eyes was similar (0.92 and 0.93, respectively). The mean rates of change in DL-AE ROI were significantly faster than for cpRNFL annulus thickness in progressing eyes (-1.28 µm/y vs. -0.83 µm/y) and nonprogressing eyes (-1.03 µm/y vs. -0.78 µm/y). Conclusions: Eye-specific ROIs identified using DL-AE analysis of OCT images show promise for improving assessment of glaucomatous progression. Translational Relevance: The detection and monitoring of structural glaucomatous progression can be improved by considering eye-specific regions of likely progression identified using deep learning.

Estimated Utility of the Short-term Assessment of Glaucoma Progression Model in Clinical Practice.

IMPORTANCE: Clinical trials of glaucoma therapies focused on protecting the optic nerve have required large sample sizes and lengthy follow-up to detect clinically relevant change due to its slow rate of progression. Whether shorter trials may be possible with more frequent testing and use of rate of change as the end point warrants further investigation. OBJECTIVE: To describe the design for the Short-term Assessment of Glaucoma Progression (STAGE) model and provide guidance on sample size and power calculations for shorter clinical trials. DESIGN, SETTING, AND PARTICIPANTS: A cohort study of patients with mild, moderate, or advanced open-angle glaucoma recruited from the Diagnostic Innovations in Glaucoma Study at the University of California, San Diego. Enrollment began in May 2012 with follow-up for every 3 months for 2 years after baseline examination. Follow-up was concluded in September 2016. Data were analyzed from July 2019 to January 2021. Visual fields (VF) and optic coherence tomography (OCT) scans were obtained at baseline and for 2 years with visits every 3 months. EXPOSURES: Glaucoma was defined as glaucomatous appearing optic discs classified by disc photographs in at least 1 eye and/or repeatable VF damage at baseline. MAIN OUTCOMES AND MEASURES: Longitudinal rates of change in retinal nerve fiber layer (RNFL) thickness and VF mean deviation (MD) are estimated in study designs of varying length and observation frequency. Power calculations as functions of study length, observation frequency, and sample size were performed. RESULTS: In a total referred sample of 97 patients with mild, moderate, or advanced glaucoma (mean [SD] age, 69 [11.4] years; 50 [51.5%] were female; 19 [19.6%]), over the 2-year follow-up, the mean VF 24-2 MD slope was -0.32 dB/y (95% CI, -0.43 to -0.21 dB/y) and the mean RNFL thickness slope was -0.54 µm/y (95% CI, -0.75 to -0.32 µm/y). Sufficient power (80%) to detect similar group differences in the rate of change in both outcomes was attained with total follow-up between 18 months and 2 years and fewer than 300 total participants. CONCLUSIONS AND RELEVANCE: In this cohort study, results from the STAGE model with reduction of the rate of progression as the end point, frequent testing, and a moderate effect size, suggest that clinical trials to test efficacy of glaucoma therapy can be completed within 18 months of follow-up and with fewer than 300 participants.

The influence of axial myopia on optic disc characteristics of glaucoma eyes.

This study characterizes differences in glaucomatous eyes with and without high axial myopia using custom automated analysis of OCT images. 452 eyes of 277 glaucoma patients were stratified into non (n = 145 eyes), mild (n = 214 eyes), and high axial myopia (axial length (AL) > 26 mm, n = 93 eyes). Optic disc ovality index, tilt and rotation angle of Bruch´s membrane opening (BMO) and peripapillary choroidal thickness (PCT) were calculated using automated and deep learning strategies. High myopic optic discs were more oval and had larger BMO tilt than mild and non-myopic discs (both p < 0.001). Mean PCT was thinnest in high myopic eyes followed by mild and non-myopic eyes (p < 0.001). BMO rotation angle, global retinal nerve fiber layer (RNFL) thickness and BMO-minimum rim width (MRW) were similar among groups. Temporal RNFL was thicker and supranasal BMO-MRW was thinner in high myopic eyes. BMO tilt and PCT showed moderate and temporal RNFL and nasal BMO-MRW showed weak but significant associations with AL in multivariable analyses (all p < 0.05). Large BMO tilt angle and thin PCT are characteristics of highly myopic discs and were not associated with severity of glaucoma. Caution should be exercised when using sectoral BMO-MRW and RNFL thickness for glaucoma management decisions in myopic eyes.

Deep Learning Estimation of 10-2 and 24-2 Visual Field Metrics Based on Thickness Maps from Macula OCT.

PURPOSE: To develop deep learning (DL) systems estimating visual function from macula-centered spectral-domain (SD) OCT images. DESIGN: Evaluation of a diagnostic technology. PARTICIPANTS: A total of 2408 10-2 visual field (VF) SD OCT pairs and 2999 24-2 VF SD OCT pairs collected from 645 healthy and glaucoma subjects (1222 eyes). METHODS: Deep learning models were trained on thickness maps from Spectralis macula SD OCT to estimate 10-2 and 24-2 VF mean deviation (MD) and pattern standard deviation (PSD). Individual and combined DL models were trained using thickness data from 6 layers (retinal nerve fiber layer [RNFL], ganglion cell layer [GCL], inner plexiform layer [IPL], ganglion cell-IPL [GCIPL], ganglion cell complex [GCC] and retina). Linear regression of mean layer thicknesses were used for comparison. MAIN OUTCOME MEASURES: Deep learning models were evaluated using R2 and mean absolute error (MAE) compared with 10-2 and 24-2 VF measurements. RESULTS: Combined DL models estimating 10-2 achieved R2 of 0.82 (95% confidence interval [CI], 0.68-0.89) for MD and 0.69 (95% CI, 0.55-0.81) for PSD and MAEs of 1.9 dB (95% CI, 1.6-2.4 dB) for MD and 1.5 dB (95% CI, 1.2-1.9 dB) for PSD. This was significantly better than mean thickness estimates for 10-2 MD (0.61 [95% CI, 0.47-0.71] and 3.0 dB [95% CI, 2.5-3.5 dB]) and 10-2 PSD (0.46 [95% CI, 0.31-0.60] and 2.3 dB [95% CI, 1.8-2.7 dB]). Combined DL models estimating 24-2 achieved R2 of 0.79 (95% CI, 0.72-0.84) for MD and 0.68 (95% CI, 0.53-0.79) for PSD and MAEs of 2.1 dB (95% CI, 1.8-2.5 dB) for MD and 1.5 dB (95% CI, 1.3-1.9 dB) for PSD. This was significantly better than mean thickness estimates for 24-2 MD (0.41 [95% CI, 0.26-0.57] and 3.4 dB [95% CI, 2.7-4.5 dB]) and 24-2 PSD (0.38 [95% CI, 0.20-0.57] and 2.4 dB [95% CI, 2.0-2.8 dB]). The GCIPL (R2 = 0.79) and GCC (R2 = 0.75) had the highest performance estimating 10-2 and 24-2 MD, respectively. CONCLUSIONS: Deep learning models improved estimates of functional loss from SD OCT imaging. Accurate estimates can help clinicians to individualize VF testing to patients.

Racial Differences in the Rate of Change in Anterior Lamina Cribrosa Surface Depth in the African Descent and Glaucoma Evaluation Study.

Purpose: The purpose of this study was to determine if the rate of change in the depth of the surface of the lamina cribrosa due to glaucomatous remodeling differs between glaucoma patients of African descent (AD) and European descent (ED). Methods: There were 1122 images taken longitudinally over an average of 3 years (range = 0.9-4.1 years) from 122 patients with glaucoma followed in the African Descent and Glaucoma Evaluation Study (ADAGES) and Diagnostic Intervention and Glaucoma Study (DIGS) were automatically segmented to compute anterior lamina cribrosa surface depth (ALCSD). The rate of ALCSD change was compared across racial groups after adjusting for baseline characteristics known to be associated with ALCSD or disease progression (visual field, ALCSD, corneal thickness, optic disk size, and age). Results: After adjusting for all other covariates, the ED group had significantly greater ALCSD posterior migration (deepening) than the AD group (difference = 2.57 µm/year, P = 0.035). There was a wider range of ALCSD change in the ED compared with the AD group, and more individuals had greater magnitude of both deepening and shallowing. No other covariates measured at baseline had independent effects on the longitudinal changes in ALCSD (baseline visual field severity, baseline ALCSD, corneal thickness, Bruch's membrane opening [BMO] area, or age). Conclusions: Glaucomatous remodeling of the lamina cribrosa differs between AD and ED patients with glaucoma. Unlike the cross-sectional associations seen with aging, in which a deeper ALCSD was seen with age in the ED group, glaucomatous remodeling in this longitudinal study resulted in more posterior migration of ALCSD in ED compared to AD patients.

Effects of Study Population, Labeling and Training on Glaucoma Detection Using Deep Learning Algorithms.

Purpose: To compare performance of independently developed deep learning algorithms for detecting glaucoma from fundus photographs and to evaluate strategies for incorporating new data into models. Methods: Two fundus photograph datasets from the Diagnostic Innovations in Glaucoma Study/African Descent and Glaucoma Evaluation Study and Matsue Red Cross Hospital were used to independently develop deep learning algorithms for detection of glaucoma at the University of California, San Diego, and the University of Tokyo. We compared three versions of the University of California, San Diego, and University of Tokyo models: original (no retraining), sequential (retraining only on new data), and combined (training on combined data). Independent datasets were used to test the algorithms. Results: The original University of California, San Diego and University of Tokyo models performed similarly (area under the receiver operating characteristic curve = 0.96 and 0.97, respectively) for detection of glaucoma in the Matsue Red Cross Hospital dataset, but not the Diagnostic Innovations in Glaucoma Study/African Descent and Glaucoma Evaluation Study data (0.79 and 0.92; P < .001), respectively. Model performance was higher when classifying moderate-to-severe compared with mild disease (area under the receiver operating characteristic curve = 0.98 and 0.91; P < .001), respectively. Models trained with the combined strategy generally had better performance across all datasets than the original strategy. Conclusions: Deep learning glaucoma detection can achieve high accuracy across diverse datasets with appropriate training strategies. Because model performance was influenced by the severity of disease, labeling, training strategies, and population characteristics, reporting accuracy stratified by relevant covariates is important for cross study comparisons. Translational Relevance: High sensitivity and specificity of deep learning algorithms for moderate-to-severe glaucoma across diverse populations suggest a role for artificial intelligence in the detection of glaucoma in primary care.

Gradient-Boosting Classifiers Combining Vessel Density and Tissue Thickness Measurements for Classifying Early to Moderate Glaucoma.

PURPOSE: To compare gradient-boosting classifier (GBC) analysis of optical coherence tomography angiography (OCTA)-measured vessel density (VD) and OCT-measured tissue thickness to standard OCTA VD and OCT thickness parameters for classifying healthy eyes and eyes with early to moderate glaucoma. DESIGN: Comparison of diagnostic tools. METHODS: A total of 180 healthy eyes and 193 glaucomatous eyes with OCTA and OCT imaging of the macula and optic nerve head (ONH) were studied. Four GBCs were evaluated that combined 1) all macula VD and thickness measurements (Macula GBC), 2) all ONH VD and thickness measurements (ONH GBC), 3) all VD measurements from the macula and ONH (vessel density GBC), and 4) all thickness measurements from the macula and ONH (thickness GBC). ROC curve (AUROC) analyses compared the diagnostic accuracy of GBCs to that of standard instrument-provided parameters. A fifth GBC that combined all parameters (full GBC) also was investigated. RESULTS: GBCs had better diagnostic accuracy than standard OCTA and OCT parameters with AUROCs ranging from 0.90 to 0.93 and 0.64 to 0.91, respectively. The full GBC (AUROC = 0.93) performed significantly better than the ONH GBC (AUROC = 0.91; P = .036) and the vessel density GBC (AUROC = 0.90; P = .010). All other GBCs performed similarly. The mean relative influence of each parameter included in the full GBC identified a combination of macular thickness and ONH VD measurements as the greatest contributors. CONCLUSIONS: GBCs that combine OCTA and OCT macula and ONH measurements can improve diagnostic accuracy for glaucoma detection compared to most but not all instrument provided parameters.