Estimating the Reliability of Glaucomatous Visual Field for the Accurate Assessment of Progression Using the Gaze-Tracking and Reliability Indices

Visual field testing in glaucoma using the Swedish Interactive Thresholding Algorithm (SITA)

The Swedish Interactive Thresholding Algorithm (SITA) is the main measurement acquisition algorithm used on the Humphrey Field Analyser, the most commonly used instrument for visual field (VF) assessment worldwide. We compare the sensitivity outputs and reliability parameters of the three currently available SITA algorithms-SITA Standard (SS), Fast (SF), and Faster (SFR), with a focus on the newly released SFR and the 24-2C test grid. SFR displays similar sensitivity outputs to SS and SF, but may not be interchangeable with SS in eyes with more severe VF loss. The reliability metric with the greatest impact on VF reliability is the level of false positives, although the recommended 15 % false positive cut off may be inappropriate as a threshold for judging whether a test is reliable and should be included for use in SFR. Finally, the 24-2C grid may be useful in flagging the presence of a clustered central VF defect, while the 10-2 grid can be used to more comprehensively characterize central field defects. We also discuss strategies to improve testing frequency in clinical practice.

Relationship between visual acuity and visual field and its reproducibility in patients with retinitis pigmentosa

BACKGROUND/OBJECTIVES

To investigate the association between visual acuity (VA) and visual field (VF) and its reproducibility in patients with retinitis pigmentosa (RP).

SUBJECTS/METHODS

The study cohort comprised 227 eyes of 227 patients with RP. The reproducibility of two Humphrey VF tests (10-2 Swedish Interactive Threshold Algorithm [SITA] tests) performed within a period of 3 months was calculated using the root mean squared error (RMSE) of each VF test point's sensitivity. The association between the logarithm of the minimum angle of resolution (logMAR) VA and VF sensitivity was investigated. Additionally, the relationship between RMSE and age, fixation loss, false positives, false negatives, and logMAR VA was determined.

RESULTS

The association between visual sensitivity and VA was most tight at the fovea, and it became weak toward the peripheral region in an eccentric manner. VF reproducibility appreciably increased as VA decreased. In particular, reproducibility was significantly decreased when logMAR VA was >0.5 compared with logMAR VA ≤ 0.

CONCLUSION

Reproducibility of VF tests decreases with a decrease in VA. Careful consideration is necessary when a patient's logMAR VA is >0.5.

The number of examinations required for the accurate prediction of the progression of the central 10-degree visual field test in glaucoma

The purpose of the study was to investigate the number of examinations required to precisely predict the future central 10-degree visual field (VF) test and to evaluate the effect of fitting non-linear models, including quadratic regression, exponential regression, logistic regression, and M-estimator robust regression model, for eyes with glaucoma. 180 eyes from 133 open angle glaucoma patients with a minimum of 13 Humphrey Field Analyzer 10-2 SITA standard VF tests were analyzed in this study. Using trend analysis with ordinary least squares linear regression (OLSLR), the first, second, and third future VFs were predicted in a point-wise (PW) manner using a varied number of prior VF sequences, and mean absolute errors (MAE) were calculated. The number of VFs needed to reach the minimum 95% confidence interval (CI) of the MAE of the OLSLR was investigated. We also examined the effect of applying other non-linear models. When predicting the first, second, and third future VFs using OLSLR, the minimum MAE was obtained using VF1-12 (2.15 ± 0.98 dB), VF1-11 (2.33 ± 1.10 dB), and VF1-10 (2.63 ± 1.36 dB), respectively. To reach the 95% CI of these MAEs, 10, 10, and 8 VFs were needed for the first, second and third future VF predictions, respectively. No improvement was observed by applying non-linear regression models. As a conclusion, approximately 8-10 VFs were needed to achieve an accurate prediction of PW VF sensitivity of the 10-degree central VF.

Real-World Analysis of the Aging Effects on Visual Field Reliability Indices in Central 10-2 Tests

We investigated the influence of aging on the reliability indices of visual field (VF) testing using a large dataset of central 10-2 program tests, including 6674 VF tests, which consisted of 1782 eyes of 1094 Japanese subjects (the mean age ± standard deviation was 66.6 ± 14.1 years). All of the combinations for each parameter, except for the pairs between age and fixation losses (FLs) or false positives (FPs) and between pattern standard deviation (PSD) and FPs, had significant correlations (p < 0.0001). Among the reliability indices, the false negatives (FNs) had the strongest correlation against age (the correlation coefficient was ρ = 0.21). Each reliability index changes differently with aging. The FLs were the highest in the first 10 s and remained constant after 20 s. The FNs remained constant for 60 s and rose steeply after 70 s. The FPs reached their highest value in 10 s and remained constant after 40 s. In mixed-effect regression analyses in 40-year-old or older subjects, older age was significantly associated with higher FNs (p < 0.0001) but not with FLs (p = 0.9014) and FPs (p = 0.9267). Compared to central 30-2 VF testing, central 10-2 VF tests were associated with smaller FLs (p < 0.0001) and FPs (p < 0.0001). In central 10-2 testing, age-related deterioration was seen in FNs but not in FLs and FPs. Choosing the 10-2 program over the 30-2 program can be effective in reducing the FL, especially in older cases with severe VF loss. This study highlighted the relationships between age and each reliability index in central 10-2 VF testing.

Gaze tracker parameters have little association with visual field metrics of intrasession frontloaded SITA-Faster 24-2 visual field results

PURPOSE

To determine the usefulness of Humphrey Field Analyser (HFA) SITA-Faster 24-2 gaze tracker outputs on interpreting intra-visit visual field (VF) result pairs.

METHODS

Analysis of 1380 right-left eye pairs and 1432 pairs of test 1-test 2 intrasession VF results of patients seen within a university-based glaucoma service was undertaken to understand gaze deviation distributions. Output gaze tracker results were aggregated into total ticks, sum of amplitudes and average amplitudes. Correlations between visual field indices (mean deviation [MD], "events" and overall hill of vision) and independent variables (age and test order) were performed using one eye from each subject.

RESULTS

There was no association of test order (right-left, test 1-test 2) with eye movements. There was a significant, but weak correlation between eye movements and age (r = 0.16). Correlations of eye movements with MD were driven by more severe MD values. There were no significant correlations between intrasession difference in eye movements and the change in MD, number of "events" and hill of vision, or in the root mean square of sensitivity and total deviation values. There was also no significant correlation between gaze tracker outputs and another commonly used "reliability" metric, false positive rate.

CONCLUSIONS

Eye movement parameters as currently reported by the HFA do not appear to be correlated with key sensitivity parameters when considering the repeatability of intrasession SITA-Faster 24-2 VF results. Thus, current gaze tracker outputs do not appear to provide clinically meaningful information for interpretation of intra-visit visual field results that cannot already be garnered using other strategies.

Association of Visual Field Pattern Reversal with Paracentral Visual Field Loss

PURPOSE

Visual field (VF) results that show more test points outside normal limits on the pattern deviation map than on the total deviation map have been assumed to be evidence of unreliable VF results. We propose the term pattern reversal to describe this VF finding and explore its association with paracentral loss.

DESIGN

Retrospective cohort and case-control studies.

PARTICIPANTS

Glaucoma and glaucoma suspect patients who completed VF testing in Veteran's Affairs ophthalmology or optometry clinics.

METHODS

In the cohort study, VF results were included that demonstrated pattern reversal. The area of pattern reversal was categorized as peripheral, paracentral, or mixed (both peripheral and paracentral). In the case-control study, a group of patients with paracentral loss confirmed on 10-2 VF tests were compared with a control group whose VF results were without paracentral loss.

MAIN OUTCOME MEASURES

In the cohort study, the calculated false-positive (FP) error rates were compared among groups categorized by area of pattern reversal. In the case-control study, the rates of pattern reversal were compared between patients with and without paracentral loss.

RESULTS

Two hundred seventeen eyes of 145 patients were included in the cohort study. Visual field results with pattern reversal and mixed loss had significantly higher FP rates compared with those with paracentral or peripheral loss only (16.25% vs. 6.26% and 8.15%, respectively; P < 0.001). Fifty-five eyes of 41 patients were included in the case group and 55 eyes of 41 patients were included in the control group. Patients with paracentral loss were more likely to have history of pattern reversal compared with those without paracentral loss (58.2% vs. 29.1%; P = 0.004). Twelve eyes with paracentral loss had 24-2 VF results that showed defects on the pattern deviation map, but not on the total deviation map.

CONCLUSIONS

Pattern reversal may be associated with paracentral VF loss and is not always associated with elevated FP rates.

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.

Faster algorithms to measure visual field using the variational Bayes linear regression model in glaucoma: comparison with SITA-Fast

AIMS

To compare the visual field (VF) test results measured with the Swedish Interactive Threshold Algorithm Fast (SITA-Fast) and newly developed variational Bayes linear regression visual field (VBLR-VF) Fast or VBLR-VF Fast+.

METHOD

Of 65 patients with glaucoma, 31 eyes of 31 patients performed VBLR-VF Fast and SITA-Fast, and 34 eyes of 34 patients performed VBLR-VF Fast+ and SITA-Fast on the same day and iterated the same procedures within 6 months using the 24-2 test grid in the current prospective study. Global index (mean deviation and pattern SD), pointwise retinal sensitivity, test duration and reliability index (fixation loss, false positive and false negative) were compared between SITA-Fast and VBLR-VF Fast or VBLR-VF Fast+.

RESULTS

Global indices were not significantly different between SITA-Fast and VBLR-VF Fast or VBLR-VF Fast+. There was no significant difference in the pointwise retinal sensitivity between the SITA-Fast and VBLR-VF Fast algorithms at the first visit, while the VBLR-VF Fast algorithm was approximately 1 dB higher compared to the SITA-Fast algorithm at the second visit. Test duration was reduced by approximately 30 s (10%) with VBLR-VF Fast and by approximately 80 s (30%) with VBLR-VF Fast+ compared with to SITA-Fast (p<0.05). Most cases showed good reliability index values; however, a marginal but significant difference was observed between the VBLR-VF and SITA-Fast algorithms.

CONCLUSION

Both VBLR-VF Fast and VBLR-VF Fast+ considerably reduced the test durations. Although there was a marginal difference in the pointwise retinal sensitivities, global indices were almost interchangeable between the VBLR-VF Fast and SITA-Fast.

Standard Reliability and Gaze Tracking Metrics in Glaucoma and Glaucoma Suspects

PURPOSE

To compare standard reliability metrics and gaze tracking (GT) metrics on the Humphrey field analyzer (HFA).

DESIGN

Retrospective cross-sectional study.

METHODS

The study was performed at the VA Medical Center, San Diego, and included 494 glaucoma and glaucoma suspect patients who had an HFA 24-2 SITA Fast visual field (VF) performed in both eyes. Standard reliability metrics (fixation loss [FL], false-positive [FP], and false-negative [FN]) were compared to GT metrics (deviations of 1°-2° [M1], deviations of 3°-5° [M3], deviations >6° [M6], and tracking failure frequency [TFF]). The main outcome measures were Spearman rank-based correlation coefficient and area under the receiver operating characteristic (AUROC) curves between standard and GT reliability metrics.

RESULTS

The 95th percentile limits for GT metrics were 66.7% for M1, 67.5% for M3, 49.5% for M6, and 79.8% for TFF. There were statistically significant correlations between standard and GT reliability metrics using the 95th percentile as a binary cutoff for GT metrics. However, low Spearman correlation values and AUROC calculations suggest little clinical significance of the associations. FN increased as VF severity worsened (P < .001). M6 was lower in eyes with mild compared to moderate and advanced VF loss (P = .012).

CONCLUSIONS

GT metrics do not have a clinically significant association with standard reliability metrics. Both FN and M6 are influenced by VF severity. Aggregate GT metrics do not aid in reliability assessment. These findings suggest that GT metrics may provide an alternative or complementary measure of VF reliability.

Computational Methods for Continuous Eye-Tracking Perimetry Based on Spatio-Temporal Integration and a Deep Recurrent Neural Network

The measurement of retinal sensitivity at different visual field locations-perimetry-is a fundamental procedure in ophthalmology. The most common technique for this scope, the Standard Automated Perimetry, suffers from several issues that make it less suitable to test specific clinical populations: it can be tedious, it requires motor manual feedback, and requires from the patient high levels of compliance. Previous studies attempted to create user-friendlier alternatives to Standard Automated Perimetry by employing eye movements reaction times as a substitute for manual responses while keeping the fixed-grid stimuli presentation typical of Standard Automated Perimetry. This approach, however, does not take advantage of the high spatial and temporal resolution enabled by the use of eye-tracking. In this study, we introduce a novel eye-tracking method to perform high-resolution perimetry. This method is based on the continuous gaze-tracking of a stimulus moving along a pseudo-random walk interleaved with saccadic jumps. We then propose two computational methods to obtain visual field maps from the continuous gaze-tracking data: the first is based on the spatio-temporal integration of ocular positional deviations using the threshold free cluster enhancement (TFCE) algorithm; the second is based on using simulated visual field defects to train a deep recurrent neural network (RNN). These two methods have complementary qualities: the TFCE is neurophysiologically plausible and its output significantly correlates with Standard Automated Perimetry performed with the Humphrey Field Analyzer, while the RNN accuracy significantly outperformed the TFCE in reconstructing the simulated scotomas but did not translate as well to the clinical data from glaucoma patients. While both of these methods require further optimization, they show the potential for a more patient-friendly alternative to Standard Automated Perimetry.

The Human Touch: Using a Webcam to Autonomously Monitor Compliance During Visual Field Assessments

Purpose

To explore the feasibility of using various easy-to-obtain biomarkers to monitor non-compliance (measurement error) during visual field assessments.

Methods

Forty-two healthy adults (42 eyes) and seven glaucoma patients (14 eyes) underwent two same-day visual field assessments. An ordinary webcam was used to compute seven potential biomarkers of task compliance, based primarily on eye gaze, head pose, and facial expression. We quantified the association between each biomarker and measurement error, as defined by (1) test–retest differences in overall test scores (mean sensitivity), and (2) failures to respond to visible stimuli on individual trials (stimuli −3 dB or more brighter than threshold).

Results

In healthy eyes, three of the seven biomarkers were significantly associated with overall (test–retest) measurement error (P = 0.003–0.007), and at least two others exhibited possible trends (P = 0.052–0.060). The weighted linear sum of all seven biomarkers was associated with overall measurement error, in both healthy eyes (r = 0.51, P < 0.001) and patients (r = 0.65, P < 0.001). Five biomarkers were each associated with failures to respond to visible stimuli on individual trials (all P < 0.001).

Conclusions

Inexpensive, autonomous measures of task compliance are associated with measurement error in visual field assessments, in terms of both the overall reliability of a test and failures to respond on particular trials (“lapses”). This could be helpful for identifying low-quality assessments and for improving assessment techniques (e.g., by discounting suspect responses or by automatically triggering comfort breaks or encouragement).

Translational Relevance

This study explores a potential way of improving the reliability of visual field assessments, a crucial but notoriously unreliable clinical measure.