Deep learning model to predict visual field in central 10° from optical coherence tomography measurement in glaucoma

Deep-Learning-Based Group Pointwise Spatial Mapping of Structure to Function in Glaucoma

Purpose

To establish generalizable pointwise spatial relationship between structure and function through occlusion analysis of a deep-learning (DL) model for predicting the visual field (VF) sensitivities from 3-dimensional (3D) OCT scan.

Design

Retrospective cross-sectional study.

Participants

A total of 2151 eyes from 1129 patients.

Methods

A DL model was trained to predict 52 VF sensitivities of 24-2 standard automated perimetry from 3D spectral-domain OCT images of the optic nerve head (ONH) with 12 915 OCT-VF pairs. Using occlusion analysis, the contribution of each individual cube covering a 240 × 240 × 31.25 μm region of the ONH to the model's prediction was systematically evaluated for each OCT-VF pair in a separate test set that consisted of 996 OCT-VF pairs. After simple translation (shifting in x- and y-axes to match the ONH center), group t-statistic maps were derived to visualize statistically significant ONH regions for each VF test point within a group. This analysis allowed for understanding the importance of each super voxel (240 × 240 × 31.25 μm covering the entire 4.32 × 4.32 × 1.125 mm ONH cube) in predicting VF test points for specific patient groups.

Main Outcome Measures

The region at the ONH corresponding to each VF test point and the effect of the former on the latter.

Results

The test set was divided to 2 groups, the healthy-to-early-glaucoma group (792 OCT-VF pairs, VF mean deviation [MD]: -1.32 ± 1.90 decibels [dB]) and the moderate-to-advanced-glaucoma group (204 OCT-VF pairs, VF MD: -17.93 ± 7.68 dB). Two-dimensional group t-statistic maps (x, y projection) were generated for both groups, assigning related ONH regions to visual field test points. The identified influential structural locations for VF sensitivity prediction at each test point aligned well with existing knowledge and understanding of structure-function spatial relationships.

Conclusions

This study successfully visualized the global trend of point-by-point spatial relationships between OCT-based structure and VF-based function without the need for prior knowledge or segmentation of OCTs. The revealed spatial correlations were consistent with previously published mappings. This presents possibilities of learning from trained machine learning models without applying any prior knowledge, potentially robust, and free from bias.

Financial Disclosures

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Abilities of circumpapillary retinal nerve fiber layer thickness and vascular density to discriminate stages in primary open-angle glaucoma

PURPOSE

To clarify the abilities of circumpapillary retinal nerve fiber layer thickness (cpRNFLT) obtained by optical coherence tomography (OCT) and circumpapillary vessel density (cpVD) measured by OCT-angiography to distinguish different stages in primary open-angle glaucoma determined by 24-2 or 30-2 static visual field (VF) testing.

METHODS

This retrospective study includes 25 healthy normal eyes of 25 subjects and 87 primary open-angle glaucoma eyes of 87 patients. Areas under the receiver operating characteristic curves (AUROC) were evaluated for determining glaucoma stages using cpRNFLT and cpVD. The absolute errors of the estimated mean total deviation (mTD) using optimal models with cpRNFLT and cpVD were also compared.

RESULTS

The AUROCs for discriminating glaucomatous eyes from normal eyes was significantly higher for cpRNFLT than the respective AUROCs for cpVD (0.969 [95% CI 0.939 to 0.998] vs. 0.872 [95% CI 0.806 to 0.938], p = 0.006), whereas cpVD had significantly higher AUROC for discriminating severe glaucoma eyes from moderate glaucoma eyes than cpRNFLT (0.771 [95% CI 0.655 to 0.886] vs. 0.578 [95% CI 0.420 to 0.736], p = 0.022). The mean absolute error in estimating mTD using both cpRNFLT and cpVD was significantly less than the error using cpRNFLT alone (4.56 ± 3.76 dB vs. 5.39 ± 4.00 dB, p = 0.027).

CONCLUSION

Our results suggest that cpVD is better for follow-ups after moderate stage. The combination of cpRNFLT and cpVD may improve VF estimation compared to cpRNFLT alone.

Predicting the Extent of Damage in the Humphrey Field Analyzer 24-2 Visual Fields Using 10-2 Test Results in Patients With Advanced Glaucoma

Purpose

To predict Humphrey Field Analyzer 24-2 test (HFA 24-2) results using 10-2 results.

Methods

A total of 175 advanced glaucoma eyes (175 patients) with HFA 24-2 mean deviation (MD24-2) of < -20 dB were prospectively followed up for five years using HFA 10-2 and 24-2 (twice and once in a year, respectively). Using all the HFA 24-2 and 10-2 test result pairs measured within three months (350 pairs from 85 eyes, training dataset), a formula to predict HFA 24-2 result using HFA 10-2 results was constructed using least absolute shrinkage and selection operator regression (LASSO). Using 90 different eyes (testing dataset), the absolute differences between the actual and LASSO-predicted MD24-2 and that between the slopes calculated using five actual and LASSO-predicted MD24-2 values, were adopted as the prediction error. Similar analyses were performed for the mean total deviation values (mTD) of the superior (or inferior) hemifield [hemi-mTDsup.24-2(-hemi-mTDinf.24-2)].

Results

The prediction error for the LASSO-predicted MD24-2 and its slope were 2.98 (standard deviation [SD] = 1.90) dB and 0.32 (0.33) dB/yr, respectively. The LASSO-predicted hemi-mTDsup.24-2 (hemi-mTDinf.24-2), and its slope were 3.02 (2.89) and 3.76 (2.72) dB, and 0.37 (0.41) and 0.44 (0.38) dB/year, respectively. These prediction errors were within two times SD of repeatability of the simulated stable HFA 24-2 VF parameter series.

Conclusions

HFA 24-2 results could be predicted using the paired HFA 10-2 results with reasonable accuracy using LASSO in patients with advanced glaucoma.

Translational Relevance

It is useful to predict HFA24-2 test from HFA10-2 test, when the former is not available, in advanced glaucoma.

Advancing Glaucoma Care: Integrating Artificial Intelligence in Diagnosis, Management, and Progression Detection

Glaucoma, the leading cause of irreversible blindness worldwide, comprises a group of progressive optic neuropathies requiring early detection and lifelong treatment to preserve vision. Artificial intelligence (AI) technologies are now demonstrating transformative potential across the spectrum of clinical glaucoma care. This review summarizes current capabilities, future outlooks, and practical translation considerations. For enhanced screening, algorithms analyzing retinal photographs and machine learning models synthesizing risk factors can identify high-risk patients needing diagnostic workup and close follow-up. To augment definitive diagnosis, deep learning techniques detect characteristic glaucomatous patterns by interpreting results from optical coherence tomography, visual field testing, fundus photography, and other ocular imaging. AI-powered platforms also enable continuous monitoring, with algorithms that analyze longitudinal data alerting physicians about rapid disease progression. By integrating predictive analytics with patient-specific parameters, AI can also guide precision medicine for individualized glaucoma treatment selections. Advances in robotic surgery and computer-based guidance demonstrate AI's potential to improve surgical outcomes and surgical training. Beyond the clinic, AI chatbots and reminder systems could provide patient education and counseling to promote medication adherence. However, thoughtful approaches to clinical integration, usability, diversity, and ethical implications remain critical to successfully implementing these emerging technologies. This review highlights AI's vast capabilities to transform glaucoma care while summarizing key achievements, future prospects, and practical considerations to progress from bench to bedside.

Deep Learning Estimation of 10-2 Visual Field Map Based on Macular Optical Coherence Tomography Angiography Measurements

PURPOSE

To develop deep learning (DL) models estimating the central visual field (VF) from optical coherence tomography angiography (OCTA) vessel density (VD) measurements.

DESIGN

Development and validation of a deep learning model.

METHODS

A total of 1051 10-2 VF OCTA pairs from healthy, glaucoma suspects, and glaucoma eyes were included. DL models were trained on en face macula VD images from OCTA to estimate 10-2 mean deviation (MD), pattern standard deviation (PSD), 68 total deviation (TD) and pattern deviation (PD) values and compared with a linear regression (LR) model with the same input. Accuracy of the models was evaluated by calculating the average mean absolute error (MAE) and the R2 (squared Pearson correlation coefficients) of the estimated and actual VF values.

RESULTS

DL models predicting 10-2 MD achieved R2 of 0.85 (95% confidence interval [CI], 74-0.92) for 10-2 MD and MAEs of 1.76 dB (95% CI, 1.39-2.17 dB) for MD. This was significantly better than mean linear estimates for 10-2 MD. The DL model outperformed the LR model for the estimation of pointwise TD values with an average MAE of 2.48 dB (95% CI, 1.99-3.02) and R2 of 0.69 (95% CI, 0.57-0.76) over all test points. The DL model outperformed the LR model for the estimation of all sectors.

CONCLUSIONS

DL models enable the estimation of VF loss from OCTA images with high accuracy. Applying DL to the OCTA images may enhance clinical decision making. It also may improve individualized patient care and risk stratification of patients who are at risk for central VF damage.

Artificial intelligence in glaucoma: opportunities, challenges, and future directions

Artificial intelligence (AI) has shown excellent diagnostic performance in detecting various complex problems related to many areas of healthcare including ophthalmology. AI diagnostic systems developed from fundus images have become state-of-the-art tools in diagnosing retinal conditions and glaucoma as well as other ocular diseases. However, designing and implementing AI models using large imaging data is challenging. In this study, we review different machine learning (ML) and deep learning (DL) techniques applied to multiple modalities of retinal data, such as fundus images and visual fields for glaucoma detection, progression assessment, staging and so on. We summarize findings and provide several taxonomies to help the reader understand the evolution of conventional and emerging AI models in glaucoma. We discuss opportunities and challenges facing AI application in glaucoma and highlight some key themes from the existing literature that may help to explore future studies. Our goal in this systematic review is to help readers and researchers to understand critical aspects of AI related to glaucoma as well as determine the necessary steps and requirements for the successful development of AI models in glaucoma.

Deep learning visual field global index prediction with optical coherence tomography parameters in glaucoma patients

The aim of this study was to predict three visual filed (VF) global indexes, mean deviation (MD), pattern standard deviation (PSD), and visual field index (VFI), from optical coherence tomography (OCT) parameters including Bruch's Membrane Opening-Minimum Rim Width (BMO-MRW) and retinal nerve fiber layer (RNFL) based on a deep-learning model. Subjects consisted of 224 eyes with Glaucoma suspects (GS), 245 eyes with early NTG, 58 eyes with moderate stage of NTG, 36 eyes with PACG, 57 eyes with PEXG, and 99 eyes with POAG. A deep neural network (DNN) algorithm was developed to predict values of VF global indexes such as MD, VFI, and PSD. To evaluate performance of the model, mean absolute error (MAE) was determined. The MAE range of the DNN model on cross validation was 1.9-2.9 (dB) for MD, 1.6-2.0 (dB) for PSD, and 5.0 to 7.0 (%) for VFI. Ranges of Pearson's correlation coefficients were 0.76-0.85, 0.74-0.82, and 0.70-0.81 for MD, PSD, and VFI, respectively. Our deep-learning model might be useful in the management of glaucoma for diagnosis and follow-up, especially in situations when immediate VF results are not available because VF test requires time and space with a subjective nature.

Improving the Accuracy and Speed of Visual Field Testing in Glaucoma With Structural Information and Deep Learning

Purpose

To assess the performance of a perimetric strategy using structure-function predictions from a deep learning (DL) model.

Methods

Visual field test-retest data from 146 eyes (75 patients) with glaucoma with (median [5th-95th percentile]) 10 [7, 10] tests per eye were used. Structure-function predictions were generated with a previously described DL model using cicumpapillary optical coherence tomography (OCT) scans. Structurally informed prior distributions were built grouping the observed measured sensitivities for each predicted value and recalculated for each subject with a leave-one-out approach. A zippy estimation by sequential testing (ZEST) strategy was used for the simulations (1000 per eye). Ground-truth sensitivities for each eye were the medians of the test-retest values. Two variations of ZEST were compared in terms of speed (average total number of presentations [NP] per eye) and accuracy (average mean absolute error [MAE] per eye), using either a combination of normal and abnormal thresholds (ZEST) or the calculated structural distributions (S-ZEST) as prior information. Two additional versions of these strategies employing spatial correlations were tested.

Results

S-ZEST was significantly faster, with a mean average NP of 213.87 (SD = 28.18), than ZEST, with a mean average NP of 255.65 (SD = 50.27) (P < 0.001). The average MAE was smaller for S-ZEST (1.98; SD = 2.37) than ZEST (2.43; SD = 2.69) (P < 0.001). Spatial correlations further improved both strategies (P < 0.001), but the differences between ZEST and S-ZEST remained significant (P < 0.001).

Conclusions

DL structure-function predictions can significantly improve perimetric tests.

Translational Relevance

DL structure-function predictions from clinically available OCT scans can improve perimetry in glaucoma patients.

Prediction of visual field progression in glaucoma: existing methods and artificial intelligence

Timely treatment is essential in the management of glaucoma. However, subjective assessment of visual field (VF) progression is not recommended, because it can be unreliable. There are two types of artificial intelligence (AI) strong and weak (machine learning). Weak AIs can perform specific tasks. Linear regression is a method of weak AI. Using linear regression in the real-world clinic has enabled analyzing and predicting VF progression. However, caution is still required when interpreting the results, because whenever the number of VF data sets investigated is small, the predictions can be inaccurate. Several other non-ordinal, or modern AI methods have been constructed to improve prediction accuracy, such as clustering and more modern AI methods of Analysis with Non-Stationary Weibull Error Regression and Spatial Enhancement (ANSWERS), Variational Bayes Linear Regression (VBLR), Kalman Filter and sparse modeling (The least absolute shrinkage and selection operator regression: Lasso). It is also possible to improve the prediction performance using retinal thickness measured with optical coherence tomography by using machine learning methods, such as multitask learning.

The application of artificial intelligence in glaucoma diagnosis and prediction

Artificial intelligence is a multidisciplinary and collaborative science, the ability of deep learning for image feature extraction and processing gives it a unique advantage in dealing with problems in ophthalmology. The deep learning system can assist ophthalmologists in diagnosing characteristic fundus lesions in glaucoma, such as retinal nerve fiber layer defects, optic nerve head damage, optic disc hemorrhage, etc. Early detection of these lesions can help delay structural damage, protect visual function, and reduce visual field damage. The development of deep learning led to the emergence of deep convolutional neural networks, which are pushing the integration of artificial intelligence with testing devices such as visual field meters, fundus imaging and optical coherence tomography to drive more rapid advances in clinical glaucoma diagnosis and prediction techniques. This article details advances in artificial intelligence combined with visual field, fundus photography, and optical coherence tomography in the field of glaucoma diagnosis and prediction, some of which are familiar and some not widely known. Then it further explores the challenges at this stage and the prospects for future clinical applications. In the future, the deep cooperation between artificial intelligence and medical technology will make the datasets and clinical application rules more standardized, and glaucoma diagnosis and prediction tools will be simplified in a single direction, which will benefit multiple ethnic groups.

Deep Learning Estimation of 10-2 Visual Field Map Based on Circumpapillary Retinal Nerve Fiber Layer Thickness Measurements

PURPOSE

To estimate central 10-degree visual field (VF) map from spectral-domain optical coherence tomography (SD-OCT) retinal nerve fiber layer thickness (RNFL) measurements in glaucoma with artificial intelligence.

DESIGN

Artificial intelligence (convolutional neural networks) study.

METHODS

This study included 5352 SD-OCT scans and 10-2 VF pairs from 1365 eyes of 724 healthy patients, patients with suspected glaucoma, and patients with glaucoma. Convolutional neural networks (CNNs) were developed to estimate the 68 individual sensitivity thresholds of 10-2 VF map using all-sectors (CNN_A) and temporal-sectors (CNN_T) RNFL thickness information of the SD-OCT circle scan (768 thickness points). 10-2 indices including pointwise total deviation (TD) values, mean deviation (MD), and pattern standard deviation (PSD) were generated using the CNN-estimated sensitivity thresholds at individual test locations. Linear regression (LR) models with the same input were used for comparison.

RESULTS

The CNN_A model achieved an average pointwise mean absolute error of 4.04 dB (95% confidence interval [CI] 3.76-4.35) and correlation coefficient (r) of 0.59 (95% CI 0.52-0.64) over 10-2 map and the mean absolute error and r of 2.88 dB (95% CI 2.63-3.15) and 0.74 (95% CI 0.67-0.80) for MD, and 2.31 dB (95% CI 2.03-2.61) and 0.59 (95% CI 0.51-0.65) for PSD estimations, respectively, significantly outperforming the LR_A model.

CONCLUSIONS

The proposed CNN_A model improved the estimation of 10-2 VF map based on circumpapillary SD-OCT RNFL thickness measurements. These artificial intelligence methods using SD-OCT structural data show promise to individualize the frequency of central VF assessment in patients with glaucoma and would enable the reallocation of resources from patients at lowest risk to those at highest risk of central VF damage.

Applications of Artificial Intelligence and Deep Learning in Glaucoma

Diagnosis and detection of progression of glaucoma remains challenging. Artificial intelligence-based tools have the potential to improve and standardize the assessment of glaucoma but development of these algorithms is difficult given the multimodal and variable nature of the diagnosis. Currently, most algorithms are focused on a single imaging modality, specifically screening and diagnosis based on fundus photos or optical coherence tomography images. Use of anterior segment optical coherence tomography and goniophotographs is limited. The majority of algorithms designed for disease progression prediction are based on visual fields. No studies in our literature search assessed the use of artificial intelligence for treatment response prediction and no studies conducted prospective testing of their algorithms. Additional challenges to the development of artificial intelligence-based tools include scarcity of data and a lack of consensus in diagnostic criteria. Although research in the use of artificial intelligence for glaucoma is promising, additional work is needed to develop clinically usable tools.

Deep learning approaches to predict 10-2 visual field from wide-field swept-source optical coherence tomography en face images in glaucoma

Close monitoring of central visual field (VF) defects with 10-2 VF helps prevent blindness in glaucoma. We aimed to develop a deep learning model to predict 10-2 VF from wide-field swept-source optical coherence tomography (SS-OCT) images. Macular ganglion cell/inner plexiform layer thickness maps with either wide-field en face images (en face model) or retinal nerve fiber layer thickness maps (RNFLT model) were extracted, combined, and preprocessed. Inception-ResNet-V2 was trained to predict 10-2 VF from combined images. Estimation performance was evaluated using mean absolute error (MAE) between actual and predicted threshold values, and the two models were compared with different input data. The training dataset comprised paired 10-2 VF and SS-OCT images of 3,025 eyes of 1,612 participants and the test dataset of 337 eyes of 186 participants. Global prediction errors (MAE_point-wise) were 3.10 and 3.17 dB for the en face and RNFLT models, respectively. The en face model performed better than the RNFLT model in superonasal and inferonasal sectors (P = 0.011 and P = 0.030). Prediction errors were smaller in the inferior versus superior hemifields for both models. The deep learning model effectively predicted 10-2 VF from wide-field SS-OCT images and might help clinicians efficiently individualize the frequency of 10-2 VF in clinical practice.

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.

Comparison of machine learning approaches for structure-function modeling in glaucoma

To evaluate machine learning (ML) approaches for structure-function modeling to estimate visual field (VF) loss in glaucoma, models from different ML approaches were trained on optical coherence tomography thickness measurements to estimate global VF mean deviation (VF MD) and focal VF loss from 24-2 standard automated perimetry. The models were compared using mean absolute errors (MAEs). Baseline MAEs were obtained from the VF values and their means. Data of 832 eyes from 569 participants were included, with 537 Asian eyes for training, and 148 Asian and 111 Caucasian eyes set aside as the respective test sets. All ML models performed significantly better than baseline. Gradient-boosted trees (XGB) achieved the lowest MAE of 3.01 (95% CI: 2.57, 3.48) dB and 3.04 (95% CI: 2.59, 3.99) dB for VF MD estimation in the Asian and Caucasian test sets, although difference between models was not significant. In focal VF estimation, XGB achieved median MAEs of 4.44 [IQR 3.45-5.17] dB and 3.87 [IQR 3.64-4.22] dB across the 24-2 VF for the Asian and Caucasian test sets and was comparable to VF estimates from support vector regression (SVR) models. VF estimates from both XGB and SVR were significantly better than the other models. These results show that XGB and SVR could potentially be used for both global and focal structure-function modeling in glaucoma.

Deep learning: applications in retinal and optic nerve diseases

Deep learning (DL) represents a paradigm-shifting, burgeoning field of research with emerging clinical applications in optometry. Unlike traditional programming, which relies on human-set specific rules, DL works by exposing the algorithm to a large amount of annotated data and allowing the software to develop its own set of rules (i.e. learn) by adjusting the parameters inside the model (network) during a training process in order to complete the task on its own. One major limitation of traditional programming is that, with complex tasks, it may require an extensive set of rules to accurately complete the assignment. Additionally, traditional programming can be susceptible to human bias from programmer experience. With the dramatic increase in the amount and the complexity of clinical data, DL has been utilised to automate data analysis and thus to assist clinicians in patient management. This review will present the latest advances in DL, for managing posterior eye diseases as well as DL-based solutions for patients with vision loss.

The Development and Clinical Application of Innovative Optical Ophthalmic Imaging Techniques

The field of ophthalmic imaging has grown substantially over the last years. Massive improvements in image processing and computer hardware have allowed the emergence of multiple imaging techniques of the eye that can transform patient care. The purpose of this review is to describe the most recent advances in eye imaging and explain how new technologies and imaging methods can be utilized in a clinical setting. The introduction of optical coherence tomography (OCT) was a revolution in eye imaging and has since become the standard of care for a plethora of conditions. Its most recent iterations, OCT angiography, and visible light OCT, as well as imaging modalities, such as fluorescent lifetime imaging ophthalmoscopy, would allow a more thorough evaluation of patients and provide additional information on disease processes. Toward that goal, the application of adaptive optics (AO) and full-field scanning to a variety of eye imaging techniques has further allowed the histologic study of single cells in the retina and anterior segment. Toward the goal of remote eye care and more accessible eye imaging, methods such as handheld OCT devices and imaging through smartphones, have emerged. Finally, incorporating artificial intelligence (AI) in eye images has the potential to become a new milestone for eye imaging while also contributing in social aspects of eye care.

Predicting Visual Fields From Optical Coherence Tomography via an Ensemble of Deep Representation Learners

PURPOSE

To develop and validate a deep learning method of predicting visual function from spectral domain optical coherence tomography (SD-OCT)-derived retinal nerve fiber layer thickness (RNFLT) measurements and corresponding SD-OCT images.

DESIGN

Development and evaluation of diagnostic technology.

METHODS

Two deep learning ensemble models to predict pointwise VF sensitivity from SD-OCT images (model 1: RNFLT profile only; model 2: RNFLT profile plus SD-OCT image) and 2 reference models were developed. All models were tested in an independent test-retest data set comprising 2181 SD-OCT/VF pairs; the median of ∼10 VFs per eye was taken as the best available estimate (BAE) of the true VF. The performance of single VFs predicting the BAE VF was also evaluated. The training data set comprised 954 eyes of 220 healthy and 332 glaucomatous participants, and the test data set, 144 eyes of 72 glaucomatous participants. The main outcome measures included the pointwise prediction mean error (ME), mean absolute error (MAE), and correlation of predictions with the BAE VF sensitivity.

RESULTS

The median mean deviation was -4.17 dB (-14.22 to 0.88). Model 2 had excellent accuracy (ME 0.5 dB, SD 0.8) and overall performance (MAE 2.3 dB, SD 3.1), and significantly (paired t test) outperformed the other methods. For single VFs predicting the BAE VF, the pointwise MAE was 1.5 dB (SD 0.7). The association between SD-OCT and single VF predictions of the BAE pointwise VF sensitivities was R² = 0.78 and R² = 0.88, respectively.

CONCLUSIONS

Our method outperformed standard statistical and deep learning approaches. Predictions of BAEs from OCT images approached the accuracy of single real VF estimates of the BAE.

Predicting Glaucoma Progression Requiring Surgery Using Clinical Free-Text Notes and Transfer Learning With Transformers

Purpose

We evaluated the use of massive transformer-based language models to predict glaucoma progression requiring surgery using ophthalmology clinical notes from electronic health records (EHRs).

Methods

Ophthalmology clinical notes for 4512 glaucoma patients at a single center from 2008 to 2020 were identified from the EHRs. Four different pre-trained Bidirectional Encoder Representations from Transformers (BERT)-based models were fine-tuned on ophthalmology clinical notes from the patients' first 120 days of follow-up for the task of predicting which patients would require glaucoma surgery. Models were evaluated with standard metrics, including area under the receiver operating characteristic curve (AUROC) and F1 score.

Results

Of the patients, 748 progressed to require glaucoma surgery (16.6%). The original BERT model had the highest AUROC (73.4%; F1 = 45.0%) for identifying these patients, followed by RoBERTa, with an AUROC of 72.4% (F1 = 44.7%); DistilBERT, with an AUROC of 70.2% (F1 = 42.5%); and BioBERT, with an AUROC of 70.1% (F1 = 41.7%). All models had higher F1 scores than an ophthalmologist's review of clinical notes (F1 = 29.9%).

Conclusions

Using transfer learning with massively pre-trained BERT-based models is a natural language processing approach that can access the wealth of clinical information stored within ophthalmology clinical notes to predict the progression of glaucoma. Future work to improve model performance can focus on integrating structured or imaging data or further tailoring the BERT models to ophthalmology domain-specific text.

Translational Relevance

Predictive models can provide the basis for clinical decision support tools to aid clinicians in identifying high- or low-risk patients to maximally tailor glaucoma treatments.

The Role of Artificial Intelligence in the Diagnosis and Management of Glaucoma

Glaucomatous optic neuropathy is the leading cause of irreversible blindness worldwide. Diagnosis and monitoring of disease involves integrating information from the clinical examination with subjective data from visual field testing and objective biometric data that includes pachymetry, corneal hysteresis, and optic nerve and retinal imaging. This intricate process is further complicated by the lack of clear definitions for the presence and progression of glaucomatous optic neuropathy, which makes it vulnerable to clinician interpretation error. Artificial intelligence (AI) and AI-enabled workflows have been proposed as a plausible solution. Applications derived from this field of computer science can improve the quality and robustness of insights obtained from clinical data that can enhance the clinician's approach to patient care. This review clarifies key terms and concepts used in AI literature, discusses the current advances of AI in glaucoma, elucidates the clinical advantages and challenges to implementing this technology, and highlights potential future applications.

Policy-Driven, Multimodal Deep Learning for Predicting Visual Fields from the Optic Disc and Optical Coherence Tomography Imaging

PURPOSE

To develop and validate a deep learning (DL) system for predicting each point on visual fields (VF) from disc and optical coherence tomography (OCT) imaging and derive a structure-function mapping.

DESIGN

Retrospective, cross-sectional database study PARTICIPANTS: 6437 patients undergoing routine care for glaucoma in three clinical sites in the UK.

METHODS

OCT and infrared reflectance (IR) optic disc imaging was paired with the closest VF within 7 days. Efficient-Net B2 was used to train two single modality DL models to predict each of the 52 sensitivity points on the 24-2 VF pattern. A policy DL model was designed and trained to fuse the two model predictions.

MAIN OUTCOME MEASURES

Pointwise Mean Absolute Error (PMAE) RESULTS: A total of 5078 imaging to VF pairs were used as a held-out test set to measure the final performance. The improvement in PMAE with the policy model was 0.485 [0.438, 0.533] dB compared to the IR image of the disc alone and 0.060 [0.047, 0.073] dB compared to the OCT alone. The improvement with the policy fusion model was statistically significant (p < 0.0001). Occlusion masking shows that the DL models learned the correct structure function mapping in a data-driven, feature agnostic fashion.

CONCLUSIONS

The multimodal, policy DL model performed the best; it provided explainable maps of its confidence in fusing data from single modalities and provides a pathway for probing the structure-function relationship in glaucoma.

A Joint Multitask Learning Model for Cross-sectional and Longitudinal Predictions of Visual Field Using OCT

Purpose

We constructed a multitask learning model (latent space linear regression and deep learning [LSLR-DL]) in which the 2 tasks of cross-sectional predictions (using OCT) of visual field (VF; central 10°) and longitudinal progression predictions of VF (30°) were performed jointly via sharing the deep learning (DL) component such that information from both tasks was used in an auxiliary manner (The Association for Computing Machinery's Special Interest Group on Knowledge Discovery and Data Mining [SIGKDD] 2021). The purpose of the current study was to investigate the prediction accuracy preparing an independent validation dataset.

Design

Cohort study.

Participants

Cross-sectional training and testing data sets included the VF (Humphrey Field Analyzer [HFA] 10-2 test) and an OCT measurement (obtained within 6 months) from 591 eyes of 351 healthy people or patients with open-angle glaucoma (OAG) and from 155 eyes of 131 patients with OAG, respectively. Longitudinal training and testing data sets included 7984 VF results (HFA 24-2 test) from 998 eyes of 592 patients with OAG and 1184 VF results (HFA 24-2 test) from 148 eyes of 84 patients with OAG, respectively. Each eye had 8 VF test results (HFA 24-2 test). The OCT sequences within the observation period were used.

Methods

Root mean square error (RMSE) was used to evaluate the accuracy of LSLR-DL for the cross-sectional prediction of VF (HFA 10-2 test). For the longitudinal prediction, the final (eighth) VF test (HFA 24-2 test) was predicted using a shorter VF series and relevant OCT images, and the RMSE was calculated. For comparison, RMSE values were calculated by applying the DL component (cross-sectional prediction) and the ordinary pointwise linear regression (longitudinal prediction).

Main Outcome Measures

Root mean square error in the cross-sectional and longitudinal predictions.

Results

Using LSLR-DL, the mean RMSE in the cross-sectional prediction was 6.4 dB and was between 4.4 dB (VF tests 1 and 2) and 3.7 dB (VF tests 1–7) in the longitudinal prediction, indicating that LSLR-DL significantly outperformed other methods.

Conclusions

The results of this study indicate that LSLR-DL is useful for both the cross-sectional prediction of VF (HFA 10-2 test) and the longitudinal progression prediction of VF (HFA 24-2 test).

Predicting 10-2 Visual Field From Optical Coherence Tomography in Glaucoma Using Deep Learning Corrected With 24-2/30-2 Visual Field

Purpose

To investigate whether a correction based on a Humphrey field analyzer (HFA) 24-2/30-2 visual field (VF) can improve the prediction performance of a deep learning model to predict the HFA 10-2 VF test from macular optical coherence tomography (OCT) measurements.

Methods

This is a multicenter, cross-sectional study. The training dataset comprised 493 eyes of 285 subjects (407, open-angle glaucoma [OAG]; 86, normative) who underwent HFA 10-2 testing and macular OCT. The independent testing dataset comprised 104 OAG eyes of 82 subjects who had undergone HFA 10-2 test, HFA 24-2/30-2 test, and macular OCT. A convolutional neural network (CNN) DL model was trained to predict threshold sensitivity (TH) values in HFA 10-2 from retinal thickness measured by macular OCT. The predicted TH values was modified by pattern-based regularization (PBR) and corrected with HFA 24-2/30-2. Absolute error (AE) of mean TH values and mean absolute error (MAE) of TH values were compared between the CNN-PBR alone model and the CNN-PBR corrected with HFA 24-2/30-2.

Results

AE of mean TH values was lower in the CNN-PBR with HFA 24-2/30-2 correction than in the CNN-PBR alone (1.9dB vs. 2.6dB; P = 0.006). MAE of TH values was lower in the CNN-PBR with correction compared to the CNN-PBR alone (4.2dB vs. 5.3 dB; P < 0.001). The inferior temporal quadrant showed lower prediction errors compared with other quadrants.

Conclusions

The performance of a DL model to predict 10-2 VF from macular OCT was improved by the correction with HFA 24-2/30-2.

Translational Relevance

This model can reduce the burden of additional HFA 10-2 by making the best use of routinely performed HFA 24-2/30-2 and macular OCT.

Telehealth and Screening Strategies in the Diagnosis and Management of Glaucoma

Telehealth has become a viable option for glaucoma screening and glaucoma monitoring due to advances in technology. The ability to measure intraocular pressure without an anesthetic and to take optic nerve photographs without pharmacologic pupillary dilation using portable equipment have allowed glaucoma screening programs to generate enough data for assessment. At home, patients can perform visual acuity testing, web-based visual field testing, rebound tonometry, and video visits with the physician to monitor for glaucomatous progression. Artificial intelligence will enhance the accuracy of data interpretation and inspire confidence in popularizing telehealth for glaucoma.

Macular Optical Coherence Tomography Imaging in Glaucoma

The advent of spectral-domain optical coherence tomography has played a transformative role in posterior segment imaging of the eye. Traditionally, images of the optic nerve head and the peripapillary area have been used to evaluate the structural changes associated with glaucoma. Recently, there is growing evidence in the literature supporting the use of macular spectral-domain optical coherence tomography as a complementary tool for clinical evaluation and research purposes in glaucoma. Containing more than 50% of retinal ganglion cells in a multilayered pattern, macula is shown to be affected even at the earliest stages of glaucomatous structural damage. Risk assessment for glaucoma progression, earlier detection of glaucomatous structural damage, monitoring of glaucoma especially in advanced cases, and glaucoma evaluation in certain ocular conditions including eyes with high myopia, positive history of disc hemorrhage, and certain optic disc phenotypes are specific domains where macular imaging yields complementary information compared to optic nerve head and peripapillary evaluation using optical coherence tomography. Moreover, the development of artificial intelligence models in data analysis has enabled a tremendous opportunity to create an integrated representation of structural and functional alterations observed in glaucoma. In this study, we aimed at providing a brief review of the main clinical applications and future potential utility of macular spectral-domain optical coherence tomography in glaucoma.

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.