Reversal of Glaucoma Hemifield Test Results and Visual Field Features in Glaucoma

Artifact Correction in Retinal Nerve Fiber Layer Thickness Maps Using Deep Learning and Its Clinical Utility in Glaucoma

Purpose

Correcting retinal nerve fiber layer thickness (RNFLT) artifacts in glaucoma with deep learning and evaluate its clinical usefulness.

Methods

We included 24,257 patients with optical coherence tomography and reliable visual field (VF) measurements within 30 days and 3,233 patients with reliable VF series of at least five measurements over ≥4 years. The artifacts are defined as RNFLT less than the known floor value of 50 µm. We selected 27,319 high-quality RNFLT maps with an artifact ratio (AR) of <2% as the ground truth. We created pseudo-artifacts from 21,722 low-quality RNFLT maps with AR of >5% and superimposed them on high-quality RNFLT maps to predict the artifact-free ground truth. We evaluated the impact of artifact correction on the structure-function relationship and progression forecasting.

Results

The mean absolute error and Pearson correlation of the artifact correction were 9.89 µm and 0.90 (P < 0.001), respectively. Artifact correction improved R2 for VF prediction in RNFLT maps with AR of >10% and AR of >20% up to 0.03 and 0.04 (P < 0.001), respectively. Artifact correction improved (P < 0.05) the AUC for progression prediction in RNFLT maps with AR of ≤10%, >10%, and >20%: (1) total deviation pointwise progression: 0.68 to 0.69, 0.62 to 0.63, and 0.62 to 0.64; and (2) mean deviation fast progression: 0.67 to 0.68, 0.54 to 0.60, and 0.45 to 0.56.

Conclusions

Artifact correction for RNFLTs improves VF and progression prediction in glaucoma.

Translational Relevance

Our model improves clinical usability of RNFLT maps with artifacts.

Combined Model of OCT Angiography and Structural OCT Parameters to Predict Paracentral Visual Field Loss in Primary Open-Angle Glaucoma

PURPOSE

To assess a model combining OCT angiography (OCTA) and OCT parameters to predict the severity of paracentral visual field (VF) loss in primary open-angle glaucoma (POAG).

DESIGN

Cross-sectional study.

PARTICIPANTS

Forty-four patients with POAG and 42 control subjects underwent OCTA and OCT imaging with a swept-source OCT device.

METHODS

The circumpapillary microvasculature was quantified for vessel density (cpVD) and flow (cpFlow) after delineation of Bruch's membrane opening and removal of large vessels. Retinal nerve fiber layer thickness (RNFLT) and Bruch's membrane opening-minimum rim width (BMO-MRW) were measured from structural OCT. Paracentral total deviation (PaTD) was defined as the average of the total deviation values within the central 10 degrees on Humphrey VF testing (24-2) for upper and lower hemifields. The OCT and OCTA parameters were measured in the affected hemisphere corresponding to the hemifield with lower PaTD for POAG patients. Models were created to predict affected PaTD based on RNFLT alone; RNFLT and BMO-MRW; OCTA alone; or RNFLT, BMO-MRW and OCTA parameters. The models were compared using coefficient of determination (r2) and Bayesian information criterion (BIC) score. Bayesian information criterion decrease of ≥6 indicates strong evidence for model improvement.

MAIN OUTCOME MEASURES

Performance of models containing OCT and OCTA parameters in predicting PaTD.

RESULTS

Patients with POAG and controls were similar in age and sex (65.9 ± 9.5 years and 38.4% male overall, P ≥ 0.56 for both). Average RNFLT, minimum RNFLT, average BMO-MRW, minimum BMO-MRW, cpVD, and cpFlow were all significantly lower (all P < 0.001) in the affected hemisphere in patients with POAG than in controls. In patients with POAG, the average mean deviation was -4.33 ± 3.25 dB; the PaTD of the affected hemifield averaged -4.55 ± 5.26 dB and correlated significantly with both OCTA and structural OCT parameters (r ≥ 0.43, P ≤ 0.004 for all). The model containing RNFLT, BMO-MRW, and OCTA parameters was superior in predicting affected PaTD (r2 = 0.47, BIC = 290.7), with higher r2 and lower BIC compared with all 3 other models.

CONCLUSIONS

A combined model of OCTA and structural OCT parameters can predict the severity of paracentral VF loss of the affected hemifield, supporting clinical utility of OCTA in patients with POAG with paracentral VF loss.

FINANCIAL DISCLOSURE(S)

Proprietary or commercial disclosure may be found after the references.

Longitudinal visual field archetypal analysis of optic neuritis treated in a clinical setting

Background/aims

We previously used archetypal analysis (AA) to create a model that quantified patterns (archetypes (ATs)) of visual field (VF) loss that can predict recovery and reveal residual VF deficits from eyes in the Optic Neuritis Treatment Trial (ONTT). We hypothesised that AA could produce similar results for ON VFs collected in clinical practice.

Methods

We applied AA to a retrospective data set of 486 VFs collected in 1 neuro-ophthalmology service from 141 eyes with acute ON and typical VF defects, to create a clinic-derived AT model. We also used the ONTT-derived AT model to analyse this new dataset. We compared the findings of both models by decomposing VFs into component ATs of varying per cent weight (PW), correlating presentation AT PW with mean deviation (MD) at final visits for each eye and identifying residual deficits in VFs considered normal.

Results

Both models, each with 16 ATs, decomposed each presentation VF into 0–6 abnormal ATs representative of known patterns of ON-related VF loss. AT1, the normal pattern in both models, correlated strongly with MD for VFs collected at presentation (r=0.82; p<0.001) and the final visit (r=0.81, p<0.001). The presentation AT1 PW was associated with improvement in MD over time. 67% of VFs considered ‘normal’ at final visit had 1.2±0.4 abnormal ATs, and both models revealed similar patterns of regional VF loss.

Conclusions

AA is a quantitative method to measure change and outcome of ON VFs. Presentation AT features are associated with MD at final visit. AA identifies residual VF deficits not otherwise indicated by MD.

Cohort Study of Race/Ethnicity and Incident Primary Open-Angle Glaucoma Characterized by Autonomously Determined Visual Field Loss Patterns

Purpose

We evaluated racial/ethnic differences in primary open-angle glaucoma (POAG) defined by machine-learning–derived regional visual field (VF) loss patterns.

Methods

Participants (N = 209,036) from the Nurses’ Health Study (NHS; 1980–2018), Nurses’ Health Study II (NHS2; 1989–2019), and Health Professionals Follow-Up Study (HPFS; 1986–2018) who were ≥40 years of age and free of glaucoma were followed biennially. Incident POAG cases (n = 1946) with reproducible VF loss were confirmed with medical records. Total deviation information from the earliest reliable glaucomatous VF for each POAG eye (n = 2564) was extracted, and machine learning analyses were used to identify optimal solutions (“archetypes”) for regional VF loss patterns. Each POAG eye was assigned a VF archetype based on the highest weighting coefficient. Multivariable-adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated using per-eye Cox proportional hazards models.

Results

We identified 14 archetypes: four representing advanced loss patterns, nine of early loss, and one of no VF loss. Compared to non-Hispanic whites, black participants had higher risk of early VF loss archetypes (HR = 1.98; 95% CI, 1.48–2.66) and even higher risk for advanced loss archetypes (HR = 6.17; 95% CI, 3.69–10.32; P-contrast = 0.0002); no differences were observed for Asians or Hispanic whites. Hispanic white participants had significantly higher risks of POAG with paracentral defects and advanced superior loss; black participants had significantly higher risks of all advanced loss archetypes and three early loss patterns, including paracentral defects.

Conclusions

Blacks, compared to non-Hispanic whites, had higher risks of POAG with early central and advanced VF loss.

Translational Relevance

In POAG, risks of VF loss regional patterns derived from machine learning algorithms showed racial differences.

Comparing Five Criteria for Evaluating Glaucomatous Visual Fields

PURPOSE

No consensus exists on the relative superiority among criteria for evaluating glaucomatous visual field (VF) damage. We compared the sensitivities and specificities of 5 criteria-Glaucoma Hemifield Test (GHT), Hoddap-Anderson-Parrish 2 (HAP2), Foster, United Kingdom Glaucoma Treatment Study (UKGTS), and Low-pressure Glaucoma Treatment Study (LoGTS)-across various levels of functional and structural glaucomatous damage.

DESIGN

Retrospective cross-sectional study.

METHODS

This single-center study included patients with suspect or known glaucoma with reliable VF (Humphrey 24-2 Swedish Interactive Thresholding Algorithm) and optical coherence tomography (OCT; Spectralis, Heidelberg Engineering) examinations within a 4-month period. One eye per patient was included. The level of functional and structural damage was defined by mean deviation (MD) and by an OCT score, respectively. We created the OCT score by counting the number of abnormal (MD percentile [P] <1%) global and sectoral averages of optic nerve head MRW, circumpapillary RNFL thickness, and macular GCL thickness. We inferred specificities and sensitivities from positive rates of the criteria in patients with low glaucomatous damage (MD at P ≥ 10% or OCT score = 0) and with higher damage (MD at P < 10% or OCT score > 0), respectively.

RESULTS

We included 1230 patients. In patients with low glaucomatous damage, HAP2 and UKGTS had higher positive rates, suggesting lower specificities, whereas GHT, Foster, and LoGTS had lower positive rates, suggesting higher specificities. In patients with higher glaucomatous damage, HAP2 and UKGTS had higher positive rates, indicating higher sensitivities, whereas GHT, Foster, and LoGTS had lower positive rates, indicating lower sensitivities.

CONCLUSIONS

No criteria had uniformly superior performance. Selection of criteria should consider the degree of damage anticipated and the desire for either higher sensitivity or specificity.

Novel technical and privacy-preserving technology for artificial intelligence in ophthalmology

PURPOSE OF REVIEW

The application of artificial intelligence (AI) in medicine and ophthalmology has experienced exponential breakthroughs in recent years in diagnosis, prognosis, and aiding clinical decision-making. The use of digital data has also heralded the need for privacy-preserving technology to protect patient confidentiality and to guard against threats such as adversarial attacks. Hence, this review aims to outline novel AI-based systems for ophthalmology use, privacy-preserving measures, potential challenges, and future directions of each.

RECENT FINDINGS

Several key AI algorithms used to improve disease detection and outcomes include: Data-driven, imagedriven, natural language processing (NLP)-driven, genomics-driven, and multimodality algorithms. However, deep learning systems are susceptible to adversarial attacks, and use of data for training models is associated with privacy concerns. Several data protection methods address these concerns in the form of blockchain technology, federated learning, and generative adversarial networks.

SUMMARY

AI-applications have vast potential to meet many eyecare needs, consequently reducing burden on scarce healthcare resources. A pertinent challenge would be to maintain data privacy and confidentiality while supporting AI endeavors, where data protection methods would need to rapidly evolve with AI technology needs. Ultimately, for AI to succeed in medicine and ophthalmology, a balance would need to be found between innovation and privacy.

Archetypal Analysis Reveals Quantifiable Patterns of Visual Field Loss in Optic Neuritis

Purpose

Identifying and monitoring visual field (VF) defects due to optic neuritis (ON) relies on qualitative clinician interpretation. Archetypal analysis (AA), a form of unsupervised machine learning, is used to quantify VF defects in glaucoma. We hypothesized that AA can identify quantifiable, ON-specific patterns (as archetypes [ATs]) of VF loss that resemble known ON VF defects.

Methods

We applied AA to a dataset of 3892 VFs prospectively collected from 456 eyes in the Optic Neuritis Treatment Trial (ONTT), and decomposed each VF into component ATs (total weight = 100%). AA of 568 VFs from 61 control eyes was used to define a minimum meaningful (≤7%) AT weight and weight change. We correlated baseline ON AT weights with global VF indices, visual acuity, and contrast sensitivity. For eyes with a dominant AT (weight ≥50%), we compared the ONTT VF classification with the AT pattern.

Results

AA generated a set of 16 ATs containing patterns seen in the ONTT. These were distinct from control ATs. Baseline study eye VFs were decomposed into 2.9 ± 1.5 ATs. AT2, a global dysfunction pattern, had the highest mean weight at baseline (36%; 95% confidence interval, 33%–40%), and showed the strongest correlation with MD (r = −0.91; P < 0.001), visual acuity (r = 0.70; P < 0.001), and contrast sensitivity (r = −0.77; P < 0.001). Of 191 baseline VFs with a dominant AT, 81% matched the descriptive classifications.

Conclusions

AA identifies and quantifies archetypal, ON-specific patterns of VF loss.

Translational Relevance

AA is a quantitative, objective method for demonstrating and monitoring change in regional VF deficits in ON.

Unsupervised Machine Learning Identifies Quantifiable Patterns of Visual Field Loss in Idiopathic Intracranial Hypertension

Purpose

Archetypal analysis, a form of unsupervised machine learning, identifies archetypal patterns within a visual field (VF) dataset such that any VF is described as a weighted sum of its archetypes (ATs) and has been used to quantify VF defects in glaucoma. We applied archetypal analysis to VFs affected by nonglaucomatous optic neuropathy caused by idiopathic intracranial hypertension (IIH).

Methods

We created an AT model from 2862 VFs prospectively collected from 330 eyes in the IIH Treatment Trial (IIHTT). We compared baseline IIH AT patterns with their descriptive VF classifications from the IIHTT.

Results

The optimum IIH AT model yielded 14 ATs resembling VF patterns reported in the IIHTT. Baseline VFs contained four or fewer meaningful ATs in 147 (89%) of study eyes. AT2 (mild general VF depression pattern) demonstrated the greatest number of study eyes with meaningful AT weight at baseline (n = 114), followed by AT1 (n = 91). Other ATs captured patterns of blind spot enlargement, hemianopia, arcuate, nasal defects, and more nonspecific patterns of general VF depression. Of all ATs, AT1 (normal pattern) had the strongest correlation with mean deviation (r = 0.69, P < 0.001). For 65 of the 93 VFs with a dominant AT, this AT matched the expert classification.

Conclusions

Archetypal analysis identifies quantifiable, archetypal VF defects that resemble those commonly seen in IIH.

Translational Relevance

Archetypal analysis provides a quantitative, objective method of measuring and monitoring disease-specific regional VF defects in IIH.

Predicting Global Test-Retest Variability of Visual Fields in Glaucoma

PURPOSE

To model the global test-retest variability of visual fields (VFs) in glaucoma.

DESIGN

Retrospective cohort study.

PARTICIPANTS

Test-retest VFs from 4044 eyes of 4044 participants.

METHODS

We selected 2 reliable VFs per eye measured with the Humphrey Field Analyzer (Swedish interactive threshold algorithm 24-2) within 30 days of each other. Each VF had fixation losses (FLs) of 33% or less, false-negative results (FNRs) of 20% or less, and false-positive results (FPRs) of 20% or less. Stepwise linear regression was applied to select the model best predicting the global test-retest variability from 3 categories of features of the first VF: (1) base parameters (age, mean deviation, pattern standard deviation, glaucoma hemifield test results, FPR, FNR, and FL); (2) total deviation (TD) at each location; and (3) computationally derived archetype VF loss patterns. The global test-retest variability was defined as root mean square deviation (RMSD) of TD values at all 52 VF locations.

MAIN OUTCOME MEASURES

Archetype models to predict the global test-retest variability.

RESULTS

The mean ± standard deviation of the root mean square deviation was 4.39 ± 2.55 dB. Between the 2 VF tests, TD values were correlated more strongly in central than in peripheral VF locations (intraclass coefficient, 0.66-0.89; P < 0.001). Compared with the model using base parameters alone (adjusted R2 = 0.45), adding TD values improved prediction accuracy of the global variability (adjusted R2 = 0.53; P < 0.001; Bayesian information criterion [BIC] decrease of 527; change of >6 represents strong improvement). Lower TD sensitivity in the outermost peripheral VF locations was predictive of higher global variability. Adding archetypes to the base model improved model performance with an adjusted R2 of 0.53 (P < 0.001) and lowering of BIC by 583. Greater variability was associated with concentric peripheral defect, temporal hemianopia, inferotemporal defect, near total loss, superior peripheral defect, and central scotoma (listed in order of decreasing statistical significance), and less normal VF results and superior paracentral defect.

CONCLUSIONS

Inclusion of archetype VF loss patterns and TD values based on first VF improved the prediction of the global test-retest variability than using traditional global VF indices alone.

Development and Comparison of Machine Learning Algorithms to Determine Visual Field Progression

Purpose

To develop and test machine learning classifiers (MLCs) for determining visual field progression.

Methods

In total, 90,713 visual fields from 13,156 eyes were included. Six different progression algorithms (linear regression of mean deviation, linear regression of the visual field index, Advanced Glaucoma Intervention Study algorithm, Collaborative Initial Glaucoma Treatment Study algorithm, pointwise linear regression [PLR], and permutation of PLR) were applied to classify each eye as progressing or stable. Six MLCs were applied (logistic regression, random forest, extreme gradient boosting, support vector classifier, convolutional neural network, fully connected neural network) using a training and testing set. For MLC input, visual fields for a given eye were divided into the first and second half and each location averaged over time within each half. Each algorithm was tested for accuracy, sensitivity, positive predictive value, and class bias with a subset of visual fields labeled by a panel of three experts from 161 eyes.

Results

MLCs had similar performance metrics as some of the conventional algorithms and ranged from 87% to 91% accurate with sensitivity ranging from 0.83 to 0.88 and specificity from 0.92 to 0.96. All conventional algorithms showed significant class bias, meaning each individual algorithm was more likely to grade uncertain cases as either progressing or stable (P ≤ 0.01). Conversely, all MLCs were balanced, meaning they were equally likely to grade uncertain cases as either progressing or stable (P ≥ 0.08).

Conclusions

MLCs showed a moderate to high level of accuracy, sensitivity, and specificity and were more balanced than conventional algorithms.

Translational Relevance

MLCs may help to determine visual field progression.

Explainable Machine Learning Model for Glaucoma Diagnosis and Its Interpretation

The aim is to develop a machine learning prediction model for the diagnosis of glaucoma and an explanation system for a specific prediction. Clinical data of the patients based on a visual field test, a retinal nerve fiber layer optical coherence tomography (RNFL OCT) test, a general examination including an intraocular pressure (IOP) measurement, and fundus photography were provided for the feature selection process. Five selected features (variables) were used to develop a machine learning prediction model. The support vector machine, C5.0, random forest, and XGboost algorithms were tested for the prediction model. The performance of the prediction models was tested with 10-fold cross-validation. Statistical charts, such as gauge, radar, and Shapley Additive Explanations (SHAP), were used to explain the prediction case. All four models achieved similarly high diagnostic performance, with accuracy values ranging from 0.903 to 0.947. The XGboost model is the best model with an accuracy of 0.947, sensitivity of 0.941, specificity of 0.950, and AUC of 0.945. Three statistical charts were established to explain the prediction based on the characteristics of the XGboost model. Higher diagnostic performance was achieved with the XGboost model. These three statistical charts can help us understand why the machine learning model produces a specific prediction result. This may be the first attempt to apply "explainable artificial intelligence" to eye disease diagnosis.

Inter-Eye Association of Visual Field Defects in Glaucoma and Its Clinical Utility

Purpose

To investigate intereye associations of visual field (VF) defects.

Methods

We selected 24-2 VF pairs of both eyes from 63,604 patients tested on the same date with mean deviation (MD) ≥ −12 dB. VFs were decomposed into one normal and 15 defect patterns previously identified using archetypal analysis. VF pattern weighting coefficients were correlated between the worse and better eyes, as defined by MD. VF defect patterns (weighting coefficients > 10%) in the better eye were predicted from weighting coefficients of the worse eye by logistic regression models, which were evaluated by area under the receiver operating characteristic curve (AUC).

Results

Intereye correlations of archetypal VF patterns were strongest for the same defect pattern between fellow eyes. The AUCs for predicting the presence of 15 defect patterns in the better eye based on the worse eye ranged from 0.69 (superior nasal step) to 0.92 (near total loss). The AUC for predicting superior paracentral loss was 0.89. Superior paracentral loss in the better eye was positively correlated with coefficients of superior paracentral loss, central scotoma, superior altitudinal defect, nasal hemianopia, and inferior paracentral loss in the worse eye, and negatively correlated with coefficients of the normal VF, superior peripheral defect, concentric peripheral defect, and temporal wedge. The parameters are presented in the descending order of statistical significance.

Conclusions

VF patterns of the worse eye are predictive of VF defects in the better eye.

Translational Relevance

Our models can potentially assist clinicians to better interpret VF loss under measurement uncertainty.

Baseline Age and Mean Deviation Affect the Rate of Glaucomatous Vision Loss

PURPOSE

The purpose of this study was to assess the relationship between the rate of the glaucomatous visual field (VF) worsening and baseline age and baseline VF mean deviation (MD).

DESIGN

This study was a retrospective, multisite cohort.

PARTICIPANTS

A total of 84,711 reliable Swedish Interactive Thresholding Algorithm 24-2 VF tests from 8167 eyes from 5644 patients with ≥6 VF tests, ≥5 years of follow-up, baseline age 18 years or above and baseline MD ≥-10 dB, and at least 2 abnormal VF tests were included from the Glaucoma Research Network Database.

METHODS

The global mean deviation rates (MDRs) and pointwise total deviation rates (TDRs) of VF progression (dB/y) were calculated for each eye using linear regression. The relationships between MDR and baseline age and MD were determined using linear mixed-effects models and logistic regression, with rapid progression defined as an MDR≤-1.0 dB/y. The relationships between TDR and baseline age and baseline MD were determined using linear mixed-effects models.

MAIN OUTCOME MEASURES

Coefficients of the regression models.

RESULTS

In individual mixed-effects models both baseline age (β=-0.0079 dB/y; P<0.001) and baseline MD (β=0.012/y; P<0.001) were associated with faster progression. All parameters were statistically significant in the full model with both parameters and their interaction (β=0.00065; P=0.0017) as covariates. With logistic regression, each year increase in baseline age increased the odds of belonging to the rapid-progressing group by a factor of 1.033, and each unit increase in baseline MD (less severe visual loss) decreased the odds by a factor of 0.8821. The mean pointwise TDR ranged from -0.21 to -0.55 dB/y, with the most rapid pointwise progression observed in the nasal and paracentral regions of the field.

CONCLUSIONS

Older age and worse MD at baseline are associated with more rapid VF progression in this large dataset. The effect of age on MDR is influenced by baseline MD severity, supporting the importance of early detection and more aggressive therapy in older patients with worse VF damage. The pointwise rate of VF loss varies across the VF, providing a means for physicians to more effectively monitor progression.

The potential application of artificial intelligence for diagnosis and management of glaucoma in adults

BACKGROUND

Glaucoma is the most frequent cause of irreversible blindness worldwide. There is no cure, but early detection and treatment can slow the progression and prevent loss of vision. It has been suggested that artificial intelligence (AI) has potential application for detection and management of glaucoma.

SOURCES OF DATA

This literature review is based on articles published in peer-reviewed journals.

AREAS OF AGREEMENT

There have been significant advances in both AI and imaging techniques that are able to identify the early signs of glaucomatous damage. Machine and deep learning algorithms show capabilities equivalent to human experts, if not superior.

AREAS OF CONTROVERSY

Concerns that the increased reliance on AI may lead to deskilling of clinicians.

GROWING POINTS

AI has potential to be used in virtual review clinics, telemedicine and as a training tool for junior doctors. Unsupervised AI techniques offer the potential of uncovering currently unrecognized patterns of disease. If this promise is fulfilled, AI may then be of use in challenging cases or where a second opinion is desirable.

AREAS TIMELY FOR DEVELOPING RESEARCH

There is a need to determine the external validity of deep learning algorithms and to better understand how the 'black box' paradigm reaches results.

An artificial intelligence model for the simulation of visual effects in patients with visual field defects

Background

This study aimed to simulate the visual field (VF) effects of patients with VF defects using deep learning and computer vision technology.

Methods

We collected 3,660 Humphrey visual fields (HVFs) as data samples, including 3,263 reliable 24-2 HVFs. The convolutional neural network (CNN) analyzed and converted the grayscale map of reliable samples into structured data. The artificial intelligence (AI) simulations were developed using computer vision technology. In statistical analyses, the pilot study determined 687 reliable samples to conduct clinical trials, and the two independent sample t-tests were used to calculate the difference of the cumulative gray values. Three volunteers evaluated the matching degree of shape and position between the grayscale map and the AI simulation, which was graded from 0 to100 scores. Based on the average ranking, the proportion of good and excellent grades was determined, and thus the reliability of the AI simulations was assessed.

Results

The reliable samples in the experimental data consisted of 1,334 normal samples and 1,929 abnormal samples. Based on the existing mature CNN model, the fully connected layer was integrated to analyze the VF damage parameters of the input images, and the prediction accuracy of the damage type of the VF defects was up to 89%. By mapping the area and damage information in the VF damage parameter quintuple data set into the real scene image and adjusting the darkening effect according to the damage parameter, the visual effects in patients were simulated in the real scene image. In the clinical validation, there was no statistically significant difference in the cumulative gray value (P>0.05). The good and excellent proportion of the average scores reached 96.0%, thus confirming the accuracy of the AI model.

Conclusions

An AI model with high accuracy was established to simulate the visual effects in patients with VF defects.

Making a Correct Diagnosis of Glaucoma: Data From the EMGT

A correct diagnosis of glaucoma established at initial visits.

Artificial Intelligence Classification of Central Visual Field Patterns in Glaucoma

PURPOSE

To quantify the central visual field (VF) loss patterns in glaucoma using artificial intelligence.

DESIGN

Retrospective study.

PARTICIPANTS

VFs of 8712 patients with 13 951 Humphrey 10-2 test results from 13 951 eyes for cross-sectional analyses, and 824 patients with at least 5 reliable 10-2 test results at 6-month intervals or more from 1191 eyes for longitudinal analyses.

METHODS

Total deviation values were used to determine the central VF patterns using the most recent 10-2 test results. A 24-2 VF within a 3-month window of the 10-2 tests was used to stage eyes into mild, moderate, or severe functional loss using the Hodapp-Anderson-Parrish scale at baseline. Archetypal analysis was applied to determine the central VF patterns. Cross-validation was performed to determine the optimal number of patterns. Stepwise regression was applied to select the optimal feature combination of global indices, average baseline decomposition coefficients from central VFs archetypes, and other factors to predict central VF mean deviation (MD) slope based on the Bayesian information criterion (BIC).

MAIN OUTCOME MEASURES

The central VF patterns stratified by severity stage based on 24-2 test results and a model to predict the central VF MD change over time using baseline test results.

RESULTS

From cross-sectional analysis, 17 distinct central VF patterns were determined for the 13 951 eyes across the spectrum of disease severity. These central VF patterns could be divided into isolated superior loss, isolated inferior loss, diffuse loss, and other loss patterns. Notably, 4 of the 5 patterns of diffuse VF loss preserved the less vulnerable inferotemporal zone, whereas they lost most of the remaining more vulnerable zone described by the Hood model. Inclusion of coefficients from central VF archetypical patterns strongly improved the prediction of central VF MD slope (BIC decrease, 35; BIC decrease of >6 indicating strong prediction improvement) than using only the global indices of 2 baseline VF results. Eyes with baseline VF results with more superonasal and inferonasal loss were more likely to show worsening MD over time.

CONCLUSIONS

We quantified central VF patterns in glaucoma, which were used to improve the prediction of central VF worsening compared with using only global indices.

Characterization of Central Visual Field Loss in End-stage Glaucoma by Unsupervised Artificial Intelligence

Importance

Although the central visual field (VF) in end-stage glaucoma may substantially vary among patients, structure-function studies and quality-of-life assessments are impeded by the lack of appropriate characterization of end-stage VF loss.

Objective

To provide a quantitative characterization and classification of central VF loss in end-stage glaucoma.

Design, Setting, and Participants

This retrospective cohort study collected data from 5 US glaucoma services from June 1, 1999, through October 1, 2014. A total of 2912 reliable 10-2 VFs of 1103 eyes from 1010 patients measured after end-stage 24-2 VFs with a mean deviation (MD) of -22 dB or less were included in the analysis. Data were analyzed from March 28, 2018, through May 23, 2019.

Main Outcomes and Measures

Central VF patterns were determined by an artificial intelligence algorithm termed archetypal analysis. Longitudinal analyses were performed to investigate whether the development of central VF defect mostly affects specific vulnerability zones.

Results

Among the 1103 patients with the most recent VFs, mean (SD) age was 70.4 (14.3) years; mean (SD) 10-2 MD, -21.5 (5.6) dB. Fourteen central VF patterns were determined, including the most common temporal sparing patterns (304 [27.5%]), followed by mostly nasal loss (280 [25.4%]), hemifield loss (169 [15.3%]), central island (120 [10.9%]), total loss (91 [8.3%]), nearly intact field (56 [5.1%]), inferonasal quadrant sparing (42 [3.8%]), and nearly total loss (41 [3.7%]). Location-specific median total deviation analyses partitioned the central VF into a more vulnerable superonasal zone and a less vulnerable inferotemporal zone. At 1-year and 2-year follow-up, new defects mostly occurred in the more vulnerable zone. Initial encroachments on an intact central VF at follow-up were more likely to be from nasal loss (11 [18.4%]; P < .001). One of the nasal loss patterns had a substantial chance at 2-year follow-up (8 [11.0%]; P = .004) to shift to total loss, whereas others did not.

Conclusions and Relevance

In this study, central VF loss in end-stage glaucoma was found to exhibit characteristic patterns that might be associated with different subtypes. Initial central VF loss is likely to be nasal loss, and 1 specific type of nasal loss is likely to develop into total loss.

New Technologies for Outcome Measures in Glaucoma: Review by the European Vision Institute Special Interest Focus Group

Glaucoma is the leading cause of irreversible blindness worldwide, with an increasing prevalence. The complexity of the disease has been a major challenge in moving the field forward with regard to both pathophysiological insight and treatment. In this context, discussing possible outcome measures in glaucoma trials is of utmost importance and clinical relevance. A recent meeting of the European Vision Institute (EVI) special interest focus group was held on "New Technologies for Outcome Measures in Retina and Glaucoma," addressing both functional and structural outcomes, as well as translational hot topics in glaucoma and retina research. In conjunction with the published literature, this review summarizes the meeting focusing on glaucoma.

The impact of artificial intelligence in the diagnosis and management of glaucoma

Deep learning (DL) is a subset of artificial intelligence (AI), which uses multilayer neural networks modelled after the mammalian visual cortex capable of synthesizing images in ways that will transform the field of glaucoma. Autonomous DL algorithms are capable of maximizing information embedded in digital fundus photographs and ocular coherence tomographs to outperform ophthalmologists in disease detection. Other unsupervised algorithms such as principal component analysis (axis learning) and archetypal analysis (corner learning) facilitate visual field interpretation and show great promise to detect functional glaucoma progression and differentiate it from non-glaucomatous changes when compared with conventional software packages. Forecasting tools such as the Kalman filter may revolutionize glaucoma management by accounting for a host of factors to set target intraocular pressure goals that preserve vision. Activation maps generated from DL algorithms that process glaucoma data have the potential to efficiently direct our attention to critical data elements embedded in high throughput data and enhance our understanding of the glaucomatous process. It is hoped that AI will realize more accurate assessment of the copious data encountered in glaucoma management, improving our understanding of the disease, preserving vision, and serving to enhance the deep bonds that patients develop with their treating physicians.

An Artificial Intelligence Approach to Detect Visual Field Progression in Glaucoma Based on Spatial Pattern Analysis

Purpose

To detect visual field (VF) progression by analyzing spatial pattern changes.

Methods

We selected 12,217 eyes from 7360 patients with at least five reliable 24-2 VFs and 5 years of follow-up with an interval of at least 6 months. VFs were decomposed into 16 archetype patterns previously derived by artificial intelligence techniques. Linear regressions were applied to the 16 archetype weights of VF series over time. We defined progression as the decrease rate of the normal archetype or any increase rate of the 15 VF defect archetypes to be outside normal limits. The archetype method was compared with mean deviation (MD) slope, Advanced Glaucoma Intervention Study (AGIS) scoring, Collaborative Initial Glaucoma Treatment Study (CIGTS) scoring, and the permutation of pointwise linear regression (PoPLR), and was validated by a subset of VFs assessed by three glaucoma specialists.

Results

In the method development cohort of 11,817 eyes, the archetype method agreed more with MD slope (kappa: 0.37) and PoPLR (0.33) than AGIS (0.12) and CIGTS (0.22). The most frequently progressed patterns included decreased normal pattern (63.7%), and increased nasal steps (16.4%), altitudinal loss (15.9%), superior-peripheral defect (12.1%), paracentral/central defects (10.5%), and near total loss (10.4%). In the clinical validation cohort of 397 eyes with 27.5% of confirmed progression, the agreement (kappa) and accuracy (mean of hit rate and correct rejection rate) of the archetype method (0.51 and 0.77) significantly (P < 0.001 for all) outperformed AGIS (0.06 and 0.52), CIGTS (0.24 and 0.59), MD slope (0.21 and 0.59), and PoPLR (0.26 and 0.60).

Conclusions

The archetype method can inform clinicians of VF progression patterns.

Machine Learning in the Detection of the Glaucomatous Disc and Visual Field

Glaucoma is the leading cause of irreversible blindness worldwide. Early detection is of utmost importance as there is abundant evidence that early treatment prevents disease progression, preserves vision, and improves patients' long-term quality of life. The structure and function thresholds that alert to the diagnosis of glaucoma can be obtained entirely via digital means, and as such, screening is well suited to benefit from artificial intelligence and specifically machine learning. This paper reviews the concepts and current literature on the use of machine learning for detection of the glaucomatous disc and visual field.

Agreement and Predictors of Discordance of 6 Visual Field Progression Algorithms

PURPOSE

To determine the agreement of 6 established visual field (VF) progression algorithms in a large dataset of VFs from multiple institutions and to determine predictors of discordance among these algorithms.

DESIGN

Retrospective longitudinal cohort study.

PARTICIPANTS

Visual fields from 5 major eye care institutions in the United States were analyzed, including a subset of eyes with at least 5 Swedish interactive threshold algorithm standard 24-2 VFs that met our reliability criteria. Of a total of 831 240 VFs, a subset of 90 713 VFs from 13 156 eyes of 8499 patients met the inclusion criteria.

METHODS

Six commonly used VF progression algorithms (mean deviation [MD] slope, VF index slope, Advanced Glaucoma Intervention Study, Collaborative Initial Glaucoma Treatment Study, pointwise linear regression, and permutation of pointwise linear regression) were applied to this cohort, and each eye was determined to be stable or progressing using each measure. Agreement between individual algorithms was tested using Cohen's κ coefficient. Bivariate and multivariate analyses were used to determine predictors of discordance (3 algorithms progressing and 3 algorithms stable).

MAIN OUTCOME MEASURES

Agreement and discordance between algorithms.

RESULTS

Individual algorithms showed poor to moderate agreement with each other when compared directly (κ range, 0.12-0.52). Based on at least 4 algorithms, 11.7% of eyes progressed. Major predictors of discordance or lack of agreement among algorithms were more depressed initial MD (P < 0.01) and older age at first available VF (P < 0.01). A greater number of VFs (P < 0.01), more years of follow-up (P < 0.01), and eye care institution (P = 0.03) also were associated with discordance.

CONCLUSIONS

This extremely large comparative series demonstrated that existing algorithms have limited agreement and that agreement varies with clinical parameters, including institution. These issues underscore the challenges to the clinical use and application of progression algorithms and of applying big-data results to individual practices.

Artificial intelligence and deep learning in ophthalmology

Artificial intelligence (AI) based on deep learning (DL) has sparked tremendous global interest in recent years. DL has been widely adopted in image recognition, speech recognition and natural language processing, but is only beginning to impact on healthcare. In ophthalmology, DL has been applied to fundus photographs, optical coherence tomography and visual fields, achieving robust classification performance in the detection of diabetic retinopathy and retinopathy of prematurity, the glaucoma-like disc, macular oedema and age-related macular degeneration. DL in ocular imaging may be used in conjunction with telemedicine as a possible solution to screen, diagnose and monitor major eye diseases for patients in primary care and community settings. Nonetheless, there are also potential challenges with DL application in ophthalmology, including clinical and technical challenges, explainability of the algorithm results, medicolegal issues, and physician and patient acceptance of the AI 'black-box' algorithms. DL could potentially revolutionise how ophthalmology is practised in the future. This review provides a summary of the state-of-the-art DL systems described for ophthalmic applications, potential challenges in clinical deployment and the path forward.

Application of Pattern Recognition Analysis to Optimize Hemifield Asymmetry Patterns for Early Detection of Glaucoma

Purpose

To assess the diagnostic utility of a new hemifield asymmetry analysis derived using pattern recognition contrast sensitivity isocontours (CSIs) within the Humphrey Field Analyzer (HFA) 24-2 visual field (VF) test grid. The performance of an optimal CSI-derived map was compared against a commercially available clustering method (Glaucoma Hemifield Test, GHT).

Methods

Five hundred VF results of 116 healthy subjects were used to determine normative distribution limits for comparisons. Pattern recognition analysis was applied to HFA 24-2 sensitivity data to determine CSI theme maps delineating clusters for hemifield comparisons. Then, 1019 VF results from 228 glaucoma patients were assessed using different clustering methods to determine the true-positive rate. We also assessed additional 354 VF results of 145 healthy subjects to determine the false-positive rate.

Results

The optimum clustering method was the CSI-derived seven-theme class map, which identified more glaucomatous VFs compared with the GHT map. The seven-class theme map also identified more cases compared with the five-, six-, and eight-class maps, suggesting no effect of number of clusters. Integrating information regarding the location of glaucomatous defects to the CSI clusters did not improve detection rate.

Conclusions

A clustering map derived using CSIs improved detection of glaucomatous VFs compared with the currently available GHT. An optimized CSI-derived map may serve as an additional means to aid earlier detection of glaucoma.

Translational Relevance

Pattern recognition–derived theme maps provide a means for guiding test point selection for asymmetry analysis in glaucoma assessment.

Development of a Visual Field Simulation Model of Longitudinal Point-Wise Sensitivity Changes From a Clinical Glaucoma Cohort

Purpose

To develop a new visual field simulation model that can recreate real-world longitudinal results at a point-wise level from a clinical glaucoma cohort.

Methods

A cohort of 367 glaucoma eyes from 265 participants seen over 10.1 ± 2.5 years were included to obtain estimates of “true” longitudinal visual field point-wise sensitivity and estimates of measurement variability. These two components were then combined to reconstruct visual field results in a manner that accounted for correlated measurement error. To determine how accurately the simulated results reflected the clinical cohort, longitudinal variability estimates of mean deviation (MD) were determined by calculating the SD of the residuals from linear regression models fitted to the MD values over time for each eye in the simulated and clinical cohorts. The new model was compared to a previous model that does not account for spatially correlated errors.

Results

The SD of all the residuals for the clinical and simulated cohorts was 1.1 dB (95% confidence interval [CI]: 1.1–1.2 dB) and 1.1 dB (95% CI: 1.1–1.1 dB), respectively, whereas it was 0.4 dB (95% CI: 0.4–0.4 dB) using the previous simulation model that did not account for correlated errors.

Conclusions

A new simulation model accounting for correlated measurement errors between visual field locations performed better than a previous model in estimating visual field variability in glaucoma.

Translational Relevance

This model can provide a powerful framework to better understand use of visual field testing in clinical practice and trials and to evaluate new methods for detecting progression.