Machine-Identified Patterns of Visual Field Loss and an Association with Rapid Progression in the Ocular Hypertension Treatment Study.

PURPOSE: To identify patterns of visual field (VF) loss based on unsupervised machine learning and to identify patterns that are associated with rapid progression. DESIGN: Cross-sectional and longitudinal study. PARTICIPANTS: A total of 2231 abnormal VFs from 205 eyes of 176 Ocular Hypertension Treatment Study (OHTS) participants followed over approximately 16 years. METHODS: Visual fields were assessed by an unsupervised deep archetypal analysis algorithm and an OHTS-certified VF reader to identify prevalent patterns of VF loss. Machine-identified patterns of glaucoma damage were compared against those patterns previously identified (expert-identified) in the OHTS in 2003. Based on the longitudinal VFs of each eye, VF loss patterns that were strongly associated with rapid glaucoma progression were identified. MAIN OUTCOME MEASURES: Machine-expert correspondence and type of patterns of VF loss associated with rapid progression. RESULTS: The average VF mean deviation (MD) at conversion to glaucoma was -2.7 decibels (dB) (standard deviation [SD] = 2.4 dB), whereas the average MD of the eyes at the last visit was -5.2 dB (SD = 5.5 dB). Fifty out of 205 eyes had MD rate of -1 dB/year or worse and were considered rapid progressors. Eighteen machine-identified patterns of VF loss were compared with expert-identified patterns, in which 13 patterns of VF loss were similar. The most prevalent expert-identified patterns included partial arcuate, paracentral, and nasal step defects, and the most prevalent machine-identified patterns included temporal wedge, partial arcuate, nasal step, and paracentral VF defects. One of the machine-identified patterns of VF loss predicted future rapid VF progression after adjustment for age, sex, and initial MD. CONCLUSIONS: An automated machine learning system can identify patterns of VF loss and could provide objective and reproducible nomenclature for characterizing early signs of visual defects and rapid progression in patients with glaucoma.

Evaluation of the Melbourne Rapid Fields Test Procedure.

SIGNIFICANCE: Both the Melbourne Rapid Fields (MRF) tablet and home versions are easy-to-use, portable, and low-cost and accurate methods of evaluating visual fields. PURPOSE: This study aimed to investigate the clinical capabilities of the MRF perimetry test by comparing it with the Humphrey Field Analyzer (HFA), determine MRF consistency, assess the influence of refractive error, ascertain ambient illumination effects, and evaluate the consistency between the tablet and Internet Web site versions of the MRF. METHODS: Forty healthy young participants with normal visual function (33 female, 7 male; average age, 24 years) underwent two MRF office-based tablet, two HFA tests, and two MRF Web site-based tests, one in our laboratory and one at home on their own computer using the 24-2 test pattern each time. An additional six healthy participants with normal visual function performed the 24-2 test with varying amounts of blur. RESULTS: The average individual sensitivity values of MRF and HFA were within 4.02 dB (right eye) and 4.15 dB (left eye). The dynamic range of the MRF was smaller (30 dB) than that of the HFA. When sensitivity values greater than 30 dB were excluded, the sensitivity differences were within 2.2 dB (right eye) and 2.46 dB (left eye) of each other. Only a small number of cases produced reliability values (false positives, false negatives, fixation losses) that were outside of normal limits. There was a high correlation between test results obtained with the tablet version of the MRF test when compared with the Internet-based Web site version. CONCLUSIONS: Quantitative visual field testing and perimetric screening procedures can be performed effectively and can provide results that are comparable with bowl perimeter test results.

A Consensus Statement on the Terminology for Automated Visual Field Abnormalities.

BACKGROUND: A multitude of terms have been used to describe automated visual field abnormalities. To date, there is no universally accepted system of definitions or guidelines. Variability among clinicians creates the risk of miscommunication and the compromise of patient care. The purposes of this study were to 1) assess the degree of consistency among a group of neuro-ophthalmologists in the description of visual field abnormalities and 2) to create a consensus statement with standardized terminology and definitions. METHODS: In phase one of the study, all neuro-ophthalmologists in Israel were asked to complete a survey in which they described the abnormalities in 10 selected automated visual field tests. In phase 2 of the study, the authors created a national consensus statement on the terminology and definitions for visual field abnormalities using a modified Delphi method. In phase 3, the neuro-ophthalmologists were asked to repeat the initial survey of the 10 visual fields using the consensus statement to formulate their answers. RESULTS: Twenty-six neuro-ophthalmologists participated in the initial survey. On average, there were 7.5 unique descriptions for each of the visual fields (SD 3.17), a description of only the location in 24.6% (SD 0.19), and an undecided response in 6.15% (SD 4.13). Twenty-two neuro-ophthalmologists participated in the creation of a consensus statement which included 24 types of abnormalities with specific definitions. Twenty-three neuro-ophthalmologists repeated the survey using the consensus statement. On average, in the repeated survey, there were 5.9 unique descriptions for each of the visual fields (SD 1.79), a description of only the location in 0.004% (SD 0.01), and an undecided response in 3.07% (SD 2.11%). Relative to the first survey, there was a significant improvement in the use of specific and decisive terminology. CONCLUSIONS: The study confirmed a great degree of variability in the use of terminology to describe automated visual field abnormalities. The creation of a consensus statement was associated with improved use of specific terminology. Future efforts may be warranted to further standardize terminology and definitions.

Do Additional Testing Locations Improve the Detection of Macular Perimetric Defects in Glaucoma?

PURPOSE: To evaluate the ability of additional central testing locations to improve detection of macular visual field (VF) defects in glaucoma. DESIGN: Prospective cross-sectional study. PARTICIPANTS: Four hundred forty healthy people and 499 patients with glaucomatous optic neuropathy (GON) were tested with a fundus tracked perimeter (CMP; CenterVue) using a 24-2 grid with 12 additional macular locations (24-2+). METHODS: Glaucomatous optic neuropathy was identified based on expert evaluation of optic nerve head photographs and OCT scans, independently of the VF. We defined macular defects as locations with measurements outside the 5% and 2% normative limits on total deviation (TD) and pattern deviation (PD) maps within the VF central 10°. Classification was based on the total number of affected macular locations (overall detection) or the largest number of affected macular locations connected in a contiguous cluster (cluster detection). Criteria based on the number of locations and cluster size were used to obtain equivalent specificity between the 24-2 grid and the 24-2+ grids, calculated using false detections in the healthy cohort. Partial areas under the receiver operating characteristic curve (pAUCs) were also compared at specificities of 95% or more. MAIN OUTCOME MEASURES: Matched specificity comparison of the ability to detect glaucomatous macular defects between the 24-2 and 24-2+ grids. RESULTS: At matched specificity, cluster detection identified more macular defects with the 24-2+ grid compared with the 24-2 grid. For example, the mean increase in percentage of detection was 8% (95% confidence interval [CI], 5%-11%) and 10% (95% CI, 7%-13%) for 5% TD and PD maps, respectively, and 5% (95% CI, 2%-7%) and 6% (95% CI, 4%-8%) for the 2% TD and PD maps, respectively. Good agreement was found between the 2 grids. The improvement measured by pAUCs was also significant but generally small. The percentage of eyes with macular defects ranged from about 30% to 50%. Test time for the 24-2+ grid was longer (21% increase) for both cohorts. Between 74% and 98% of defects missed by the 24-2 grid had at least 1 location with sensitivity of < 20 dB. CONCLUSIONS: Visual field examinations with additional macular locations can improve the detection of macular defects in GON modestly without loss of specificity when appropriate criteria are selected.

A Comparison between the Compass Fundus Perimeter and the Humphrey Field Analyzer.

PURPOSE: To evaluate relative diagnostic precision and test-retest variability of 2 devices, the Compass (CMP, CenterVue, Padova, Italy) fundus perimeter and the Humphrey Field Analyzer (HFA, Zeiss, Dublin, CA), in detecting glaucomatous optic neuropathy (GON). DESIGN: Multicenter, cross-sectional, case-control study. PARTICIPANTS: We sequentially enrolled 499 patients with glaucoma and 444 normal subjects to analyze relative precision. A separate group of 44 patients with glaucoma and 54 normal subjects was analyzed to assess test-retest variability. METHODS: One eye of recruited subjects was tested with the index tests: HFA (Swedish interactive thresholding algorithm [SITA] standard strategy) and CMP (Zippy Estimation by Sequential Testing [ZEST] strategy), 24-2 grid. The reference test for GON was specialist evaluation of fundus photographs or OCT, independent of the visual field (VF). For both devices, linear regression was used to calculate the sensitivity decrease with age in the normal group to compute pointwise total deviation (TD) values and mean deviation (MD). We derived 5% and 1% pointwise normative limits. The MD and the total number of TD values below 5% (TD 5%) or 1% (TD 1%) limits per field were used as classifiers. MAIN OUTCOME MEASURES: We used partial receiver operating characteristic (pROC) curves and partial area under the curve (pAUC) to compare the diagnostic precision of the devices. Pointwise mean absolute deviation and Bland-Altman plots for the mean sensitivity (MS) were computed to assess test-retest variability. RESULTS: Retinal sensitivity was generally lower with CMP, with an average mean difference of 1.85±0.06 decibels (dB) (mean ± standard error, P < 0.001) in healthy subjects and 1.46±0.05 dB (mean ± standard error, P < 0.001) in patients with glaucoma. Both devices showed similar discriminative power. The MD metric had marginally better discrimination with CMP (pAUC difference ± standard error, 0.019±0.009, P = 0.035). The 95% limits of agreement for the MS were reduced by 13% in CMP compared with HFA in participants with glaucoma and by 49% in normal participants. Mean absolute deviation was similar, with no significant differences. CONCLUSIONS: Relative diagnostic precision of the 2 devices is equivalent. Test-retest variability of MS for CMP was better than for HFA.

Oral Memantine for the Treatment of Glaucoma: Design and Results of 2 Randomized, Placebo-Controlled, Phase 3 Studies.

PURPOSE: To evaluate the effectiveness and safety of oral memantine as a potential neuroprotective agent in open-angle glaucoma (OAG) at risk for progression. DESIGN: Two randomized, double-masked, placebo-controlled, parallel-group, multicenter, 48-month studies identically designed, initiated 1 year apart, and completed in 2006. Protocol amendments included a 1-year extension (first study) and change in primary endpoint and analysis (second study). PARTICIPANTS: Patients (2298 total) with bilateral OAG; glaucomatous optic disc damage and visual field loss in 1 eye; glaucomatous optic disc damage and/or visual field loss in the contralateral eye (at screening), topically treated or untreated intraocular pressure (IOP) of 21 mmHg or less (at baseline); and at risk of glaucomatous progression (per prespecified criteria). METHODS: Patients were randomized 3:2:2 to receive memantine 20 mg, memantine 10 mg, or placebo tablets daily. Glaucomatous progression was assessed in the intent-to-treat population by full-threshold standard automated perimetry (SAP), frequency doubling technology (FDT), and stereoscopic optic disc photographs, standardized by quality control assessment at centralized reading centers. Safety evaluations included adverse events (AEs), best-corrected visual acuity, biomicroscopy, IOP, and ophthalmoscopy. Efficacy data from each study were analyzed per protocol. Pooled analyses of efficacy and safety data were also performed. MAIN OUTCOME MEASURES: The predefined primary efficacy measure was glaucomatous visual field progression, as measured by SAP. Additional efficacy measures included glaucomatous progression of visual field (FDT) and optic nerve damage (stereoscopic optic disc photographs). RESULTS: The proportion of patients who completed the studies was similar among groups (80%-83%). Compared with placebo, daily treatment with memantine 10 mg or 20 mg for 48 months did not delay glaucomatous progression significantly in the individual studies and pooled analyses. The pooled risk reduction ratio (95% confidence interval) assessed by SAP was -0.13 (-0.40, 0.09) and -0.17 (-0.46, 0.07) for memantine 10 mg and 20 mg, respectively. Results were similar per FDT and stereoscopic optic disc photographs. The most common AEs leading to treatment discontinuations were dizziness, headache, fatigue, and nausea. CONCLUSIONS: With technologies available when the studies were conducted, daily treatment with memantine over 48 months was not shown to prevent glaucomatous progression in this patient population.

Predicting conversion to glaucoma using standard automated perimetry and frequency doubling technology.

PURPOSE: To test the hypothesis that development of glaucomatous visual fields can be predicted several years earlier from prior visual field information. METHODS: One-hundred and seven eyes with glaucomatous optic neuropathy (n = 47 eyes) or which were suspicious for glaucoma (n = 60) were prospectively enrolled in a longitudinal study. Visual fields were evaluated on an annual basis using standard automated perimetry (SAP), the original version of frequency doubling technology (FDT) perimetry, and a custom version of FDT that used the 24-2 stimulus pattern. All SAP fields were within normal limits at the initial visit. When the SAP glaucoma hemifield test was 'outside normal limits' or the pattern standard deviation probability was worse than the lower 5th percentile or more than two clustered locations at the p < 0.05 level were present on the pattern deviation probability plot, an eye was defined as being abnormal. We used a classification tree analysis to predict which eyes would convert, using only baseline test results. RESULTS: Classification trees that were constructed using only baseline data had excellent specificity (near 100%) but worse sensitivity (25-50%) for predicting which eyes would convert during follow-up. CONCLUSIONS: Predictive information is present in visual field results, even when they are still within normal limits.

Retinal structure and function in vigabatrin-treated adult patients with refractory complex partial seizures.

OBJECTIVE: Evaluate visual-field and retinal-structure changes following adjunctive vigabatrin treatment in vigabatrin-naive adults with refractory complex partial seizures (rCPS). METHODS: Prospective, longitudinal, single-arm, open-label study (NCT01278173). Eligible patients (≥2 seizures/month who failed ≥3 therapies) who could reliably perform perimetry (Humphrey automated static) and retinal-structure assessment (spectral-domain optical coherence tomography) prior to vigabatrin exposure. Following vigabatrin initiation, testing occurred within 1 month (reference) and 3, 6, 9, and 12 months. End points included mean change from reference in mean deviation (dB) and average retinal nerve fiber layer (RNFL) thickness, visual-acuity changes from baseline, and number of patients who met predefined vision-parameter changes at two (confirmed) or three (persistent) consecutive visits. RESULTS: Sixty-five of 91 screened patients received ≥1 vigabatrin dose (all-patients-treated set [APTS]); 55 had valid reference and ≥1 post-reference assessments (full-analysis set [FAS]). Thirty-six APTS patients with valid pre-/post-reference values completed all planned visits (per-protocol set [PPS]). Thirty-eight (59%) APTS patients completed the study; 27 (42%) withdrew (none for visual-field changes); 32% and 15% had abnormally thin RNFL and abnormal visual acuity at baseline, respectively; 20% had abnormal central 30 degree visual fields in the reference period. No significant mean near visual-field changes were observed (PPS); mean change in average RNFL thickness increased significantly (1-year data: Left-eye: 6.37 µm, confidence interval (CI) 4.66-8.09; right-eye: 7.24 µm CI 5.47-9.01; PPS). No confirmed three-line decreases in visual acuity (FAS) were observed; five patients had predefined confirmed/persistent visual-field changes (FAS). All vision-related adverse events were nonserious; the most common was vision blurred (9%). SIGNIFICANCE: Prior to vigabatrin initiation, rCPS patients may already exhibit vision deficits. Up to 1 year of adjunctive vigabatrin treatment did not significantly change population near visual fields. Five patients met predefined visual-field-change criteria. RNFL thickening of unknown clinical significance was observed. Limitations include single-arm, open-label design; patients' inability to perform ophthalmic/visual-field examinations; and limited vigabatrin-exposure duration.

Factors Affecting Visual Field Outcomes in the Idiopathic Intracranial Hypertension Treatment Trial.

BACKGROUND: To determine the prevalence of visual field (VF) performance failures (PF) and treatment failures (TFs), and identify factors associated with PFs in the Idiopathic Intracranial Hypertension Treatment Trial (IIHTT). METHODS: A total of 165 participants from 38 sites with idiopathic intracranial hypertension (IIH) and mild visual loss were randomized to either acetazolamide-plus diet or placebo-plus diet. The IIHTT Visual Field Reading Center evaluated 2950 Swedish Interactive Threshold Algorithm Standard 24-2 VFs from the enrolled participants. A TF was defined when the participant's VF mean deviation (MD) worsened ≥2 to 3 dB from the average baseline MD (range of -2 to -7 dB) with a second retest confirming the visual deterioration. A PF was determined when the participant's: 1) VF results met TF criteria but were not confirmed on retest, 2) deterioration was confirmed on retest but the IIHTT Adjudication Committee concluded a TF was clinically unlikely. RESULTS: TF was detected in 7/165 (4%) of the participants and PF was detected in 35/165 (21%) of the participants on at least 1 examination. Four of the 35 PFs were adjudicated for TF, however based on clinical review by the adjudication committee and a third retest, they were judged as PFs. Of the 2,950 total IIHTT VF examinations, 2.7% met PF criteria. CONCLUSIONS: PF was confirmed in 21% of subjects and in 2.7% of the total number of VF examinations and was reversible on repeat testing. We recommend retesting when perimetric worsening occurs in otherwise clinically stable or improving IIH patients.

A deep learning model incorporating spatial and temporal information successfully detects visual field worsening using a consensus based approach.

Glaucoma is a leading cause of irreversible blindness, and its worsening is most often monitored with visual field (VF) testing. Deep learning models (DLM) may help identify VF worsening consistently and reproducibly. In this study, we developed and investigated the performance of a DLM on a large population of glaucoma patients. We included 5099 patients (8705 eyes) seen at one institute from June 1990 to June 2020 that had VF testing as well as clinician assessment of VF worsening. Since there is no gold standard to identify VF worsening, we used a consensus of six commonly used algorithmic methods which include global regressions as well as point-wise change in the VFs. We used the consensus decision as a reference standard to train/test the DLM and evaluate clinician performance. 80%, 10%, and 10% of patients were included in training, validation, and test sets, respectively. Of the 873 eyes in the test set, 309 [60.6%] were from females and the median age was 62.4; (IQR 54.8-68.9). The DLM achieved an AUC of 0.94 (95% CI 0.93-0.99). Even after removing the 6 most recent VFs, providing fewer data points to the model, the DLM successfully identified worsening with an AUC of 0.78 (95% CI 0.72-0.84). Clinician assessment of worsening (based on documentation from the health record at the time of the final VF in each eye) had an AUC of 0.64 (95% CI 0.63-0.66). Both the DLM and clinician performed worse when the initial disease was more severe. This data shows that a DLM trained on a consensus of methods to define worsening successfully identified VF worsening and could help guide clinicians during routine clinical care.

Mammalian fatty acid synthase activity from crude tissue lysates tracing ¹³C-labeled substrates using gas chromatography-mass spectrometry.

Fatty acid synthase (FASN or FAS, EC 2.3.1.85) is the sole mammalian enzyme to synthesize fatty acids de novo from acetyl- and malonyl-coenzyme A (CoA) esters. This article describes a new method that directly quantifies uniformly labeled (¹³C16-labeled palmitate ([¹³C16]palmitate) by tracing [¹³C2]acetyl-CoA and [¹³C3]malonyl-CoA using an in vitro FASN assay. This method used gas chromatography-mass spectrometry (GC-MS) to detect [¹³C16]palmitate carboxylate anions (m/z 271) of pentafluorobenzyl (PFB) derivatives and was highly sensitive at femtomole quantities. Uniformly incorporated [¹³C16]palmitate was the primary product of both recombinant and crude tissue lysate FASN. Quantification of FASN protein within crude tissue lysates ensured equal FASN amounts, preserved steady-state kinetics, and enabled calculation of FASN-specific activity. FASN activity determined by [¹³C16]palmitate synthesis was consistent with values obtained from ß-nicotinamide adenine dinucleotide 2'-phosphate (NADPH) oxidation assays. Analysis of FASN activity from tissue extracts was not hampered by contaminating enzymes or preexisting fatty acids. Crude mammary gland and liver lysates had significantly different activities at 82 and 65 nmol min⁻¹ mg⁻¹, respectively, suggesting that tissue-specific activity levels differ in a manner unrelated to FASN amount. GC-MS quantification of [¹³C16]palmitate synthesis permits sensitive evaluation of FASN activity from tissues of varied physiological states and of purified FASN activity in the presence of modifying proteins, enzymes, or drugs.

The effect of test variability on the structure-function relationship in early glaucoma.

PURPOSE: To determine whether the weakness of the structure-function relationship could be produced by test variability alone, without implying underlying dissociation between the true rates of structural and functional change. METHODS: Perimetric mean deviation (MD), and rim area (RA) and cup volume (CV) from confocal scanning laser ophthalmoscopy, over six visits, were taken from 166 eyes of 92 participants with high-risk ocular hypertension or suspected/early glaucoma in the Portland Progression Project. Models were created of each measure's variability. A further model predicted the rate of functional change from the rate of structural change. These were used to generate realistic simulated sequences of both functional and structural data with different standard deviations σ between the underlying rates of change. 'Observed' structure-function relationships were calculated. An empirical p-value was derived, equaling the proportion of simulated series for which the 'observed' structure-function dissociation was greater than that seen in patient data. RESULTS: The correlation between the rates of structural (RA) and functional (MD) change was 0.171, consistent with σ < 0.02 dB/yr. Using CV, the correlation was -0.091, consistent with σ < 0.01 dB/yr. By comparison, the models predicted that the standard deviation of the rate of functional change for a healthy eye due to test variability would be 0.18 dB/yr. CONCLUSION: Test variability is sufficiently large that realistic patient data can be simulated without requiring a large variability between the underlying rates of structural and functional change. This absence of implied dissociation is a necessary condition for it to be valid to combine structural and functional measures to improve estimates of functional change and/or to reduce perimetric variability.

An Objective and Easy-to-Use Glaucoma Functional Severity Staging System Based on Artificial Intelligence.

OBJECTIVE: The objective of this study was to develop an objective and easy-to-use glaucoma staging system based on visual fields (VFs). SUBJECTS AND PARTICIPANTS: A total of 13,231 VFs from 8077 subjects were used to develop models and 8024 VFs from 4445 subjects were used to validate models. METHODS: We developed an unsupervised machine learning model to identify clusters with similar VF values. We annotated the clusters based on their respective mean deviation (MD). We computed optimal MD thresholds that discriminate clusters with the highest accuracy based on Bayes minimum error principle. We evaluated the accuracy of the staging system and validated findings based on an independent validation dataset. RESULTS: The unsupervised k -means algorithm discovered 4 clusters with 6784, 4034, 1541, and 872 VFs and average MDs of 0.0 dB (±1.4: SD), -4.8 dB (±1.9), -12.2 dB (±2.9), and -23.0 dB (±3.8), respectively. The supervised Bayes minimum error classifier identified optimal MD thresholds of -2.2, -8.0, and -17.3 dB for discriminating normal eyes and eyes at the early, moderate, and advanced stages of glaucoma. The accuracy of the glaucoma staging system was 94%, based on identified MD thresholds with respect to the initial k -means clusters. CONCLUSIONS: We discovered that 4 severity levels based on MD thresholds of -2.2, -8.0, and -17.3 dB, provides the optimal number of severity stages based on unsupervised and supervised machine learning. This glaucoma staging system is unbiased, objective, easy-to-use, and consistent, which makes it highly suitable for use in glaucoma research and for day-to-day clinical practice.

Detecting glaucoma from multi-modal data using probabilistic deep learning.

Objective: To assess the accuracy of probabilistic deep learning models to discriminate normal eyes and eyes with glaucoma from fundus photographs and visual fields. Design: Algorithm development for discriminating normal and glaucoma eyes using data from multicenter, cross-sectional, case-control study. Subjects and participants: Fundus photograph and visual field data from 1,655 eyes of 929 normal and glaucoma subjects to develop and test deep learning models and an independent group of 196 eyes of 98 normal and glaucoma patients to validate deep learning models. Main outcome measures: Accuracy and area under the receiver-operating characteristic curve (AUC). Methods: Fundus photographs and OCT images were carefully examined by clinicians to identify glaucomatous optic neuropathy (GON). When GON was detected by the reader, the finding was further evaluated by another clinician. Three probabilistic deep convolutional neural network (CNN) models were developed using 1,655 fundus photographs, 1,655 visual fields, and 1,655 pairs of fundus photographs and visual fields collected from Compass instruments. Deep learning models were trained and tested using 80% of fundus photographs and visual fields for training set and 20% of the data for testing set. Models were further validated using an independent validation dataset. The performance of the probabilistic deep learning model was compared with that of the corresponding deterministic CNN model. Results: The AUC of the deep learning model in detecting glaucoma from fundus photographs, visual fields, and combined modalities using development dataset were 0.90 (95% confidence interval: 0.89-0.92), 0.89 (0.88-0.91), and 0.94 (0.92-0.96), respectively. The AUC of the deep learning model in detecting glaucoma from fundus photographs, visual fields, and both modalities using the independent validation dataset were 0.94 (0.92-0.95), 0.98 (0.98-0.99), and 0.98 (0.98-0.99), respectively. The AUC of the deep learning model in detecting glaucoma from fundus photographs, visual fields, and both modalities using an early glaucoma subset were 0.90 (0.88,0.91), 0.74 (0.73,0.75), 0.91 (0.89,0.93), respectively. Eyes that were misclassified had significantly higher uncertainty in likelihood of diagnosis compared to eyes that were classified correctly. The uncertainty level of the correctly classified eyes is much lower in the combined model compared to the model based on visual fields only. The AUCs of the deterministic CNN model using fundus images, visual field, and combined modalities based on the development dataset were 0.87 (0.85,0.90), 0.88 (0.84,0.91), and 0.91 (0.89,0.94), and the AUCs based on the independent validation dataset were 0.91 (0.89,0.93), 0.97 (0.95,0.99), and 0.97 (0.96,0.99), respectively, while the AUCs based on an early glaucoma subset were 0.88 (0.86,0.91), 0.75 (0.73,0.77), and 0.92 (0.89,0.95), respectively. Conclusion and relevance: Probabilistic deep learning models can detect glaucoma from multi-modal data with high accuracy. Our findings suggest that models based on combined visual field and fundus photograph modalities detects glaucoma with higher accuracy. While probabilistic and deterministic CNN models provided similar performance, probabilistic models generate certainty level of the outcome thus providing another level of confidence in decision making.

Perimetric indices as predictors of future glaucomatous functional change.

PURPOSE: To evaluate the use of global indices summarizing the current status of a patient's visual field as predictors of their future rate of change. METHODS: Ninety-five subjects with early or suspected glaucoma were studied, of whom 50 exhibited glaucomatous optic neuropathy (GON) at baseline. Subjects underwent seven annual visual field tests. Results from the first test in the sequence were used to predict their subsequent rate of change. Two global indices were considered: mean deviation (MD) and pattern standard deviation (PSD). RESULTS: Using multiple linear regression, baseline MD predicted subsequent slope of MD significantly better than baseline PSD predicted subsequent slope of PSD (p = 0.017). Using multiple logistic regression, a worse initial MD was predictive of being in the worst tertile for subsequent change in MD (pseudo-R2 = 0.33 for subjects with GON at baseline and 0.31 for those without). Worse initial PSD was not predictive of being in the worst tertile for subsequent change in PSD (pseudo-R2 = 0.09 with GON at baseline, 0.10 without). CONCLUSIONS: Among patients with otherwise similar clinical profiles, a worse visual field at baseline, as measured by the global index MD, is predictive of a more rapid future rate of change. This should be taken into account when clinical decisions are made concerning management of patients who already have some visual field damage at presentation.

Cup size predicts subsequent functional change in early glaucoma.

PURPOSE: To identify structural predictors of functional change in patients with early glaucoma or risk factors for development of glaucoma. METHODS: One hundred twenty-nine participants with either high-risk ocular hypertension, suspected, or early glaucoma were followed for 10 annual visits. Standard Automated Perimetry was performed at each visit, along with Confocal Scanning Laser Ophthalmoscopy (CSLO). Mean deviation (MD) at the same visit and the subsequent rate of change in MD were predicted by linear regression models based on CSLO parameters along with intraocular pressure, central corneal thickness, and treatment status. Models were assessed by comparing the correlations between predicted and observed perimetric results. RESULTS: The model using rim area predicted MD at the same visit with correlation 0.445. Using cup area, the equivalent correlation was r = 0.411, lower than the model using rim area with p = 0.096. Using cup volume, the correlation was r = 0.436, and the comparison had p = 0.714; using disc area r = 0.396 and p = 0.011. The model using rim area to predict the subsequent rate of change of MD had r = 0.539. Models using cup area (r = 0.631), cup volume (r = 0.649), or disc area (r = 0.602) were significantly better; p = 0.003, p = 0.001, and p = 0.044, respectively. CONCLUSIONS: Lower rim area, as measured by CSLO, is predictive of worse function when measured on the same date in participants with high-risk ocular hypertension, suspected, and early glaucoma. However, cup area, cup volume, and disc area may all be more predictive of the rate of subsequent functional change. Parameters should be chosen based on the specific purpose intended and questions asked, rather than using the same parameters in all circumstances.

A history of perimetry and visual field testing.

Perimetry and visual field testing have been used as clinical ophthalmic diagnostic tools for many years, and this manuscript will provide a brief historical overview of these procedures and the individuals who developed them. Today, we have many different forms of perimetry that are designed to evaluate different locations within the visual pathways and various mechanisms and subsets of mechanisms within the visual system. However, the most widely used method of performing perimetry and visual field testing has not substantially changed for more than 150 years, consisting of detecting a small target superimposed on a uniform background at different locations within the field of view. Although the basic test procedure has remained similar throughout the ages, there have been many advances in test administration, standardization, statistical evaluation, clinical analysis, interpretation, and prediction of outcome based on visual field findings.

Variability of rarebit and standard perimetry sizes I and III in normals.

PURPOSE: To compare test-retest variability in rarebit perimetry (RBP) and conventional standard automated perimetry (SAP) using Goldmann size I and III stimuli. METHODS: Seventeen normal subjects underwent RBP, size I SAP, and size III SAP on five separate visits within a 5-week period. Test-retest variability could not be directly compared because of differences in units. Instead, data were transformed to lie on the same scale by calculating a percentile range to mean ratio for each test. RESULTS: Percentile range to mean ratio was highest for size I (3.42 ± 0.62), followed by size III (2.29 ± 0.55), and RBP (0.29 ± 0.10). Results for all three tests were significantly different from each other. CONCLUSIONS: Comparison of tests with different scales is problematic, but RBP's retest variability appears to be significantly less than both sizes I and III SAP in the measurement of normal subjects.

Measuring visual function in age-related macular degeneration with frequency-doubling (matrix) perimetry.

PURPOSE: To determine the agreement between the Humphrey Matrix perimeter 10-2 test and the 10-2 Humphrey Field Analyzer (HFA) test when assessing visual function in patients with age-related macular degeneration (AMD). METHODS: Forty-two eyes of 42 subjects with AMD (average 75.0 years, SD=6.2: median visual acuity in logarithm of the minimum angle of resolution of 0.26, range, -0.12 to 1.04) were evaluated with the Matrix and HFA 10-2 visual field tests. Mean deviation (MD), pattern standard deviation, and test time were recorded. We calculated spatial concordance of individual test locations, being the proportion of spatially agreeing locations with identical classification (normal vs. abnormal, p < 5%) on the pattern deviation plot. As multiple HFA stimuli overlapped with some Matrix locations, several criteria for grouping HFA data into locations were investigated. RESULTS: Both MD and pattern standard deviation were significantly correlated for the two devices (r(2)=0.79 and r(2)=0.80, respectively, p<0.0001). Using our standard criterion for abnormal HFA locations (≥50% stimuli abnormal), the median spatial concordance was 0.76, with 95% of tests giving a concordance of ≥0.59. A small, but significant, increase in concordance occurred when a stricter criterion (all stimuli abnormal at a location) was applied. Median fixation loss percentages were 7 and 0% for the HFA and Matrix, respectively. Visual acuity in logarithm of the minimum angle of resolution showed modest correlations with both defect depth (HFA MD: r(2)=0.39, p<0.0001) and size of defect (number of abnormal points on the HFA: r(2)=0.24, p<0.0001). CONCLUSIONS: Using a simple metric to calculate spatial concordance, the Matrix 10-2 test quantifies the spatial extent of significant depression of the central visual fields in AMD in a manner similar to the HFA 10-2. The spatial extent and depth of central visual field loss in AMD are only modestly predicted by visual acuity measurements.