Evaluation of Age-related Macular Degeneration With Optical Coherence Tomography

Periodontitis and Outer Retinal Thickness: a Cross-Sectional Analysis of the United Kingdom Biobank Cohort

Purpose

Periodontitis, a ubiquitous severe gum disease affecting the teeth and surrounding alveolar bone, can heighten systemic inflammation. We investigated the association between very severe periodontitis and early biomarkers of age-related macular degeneration (AMD), in individuals with no eye disease.

Design

Cross-sectional analysis of the prospective community-based cohort United Kingdom (UK) Biobank.

Participants

Sixty-seven thousand three hundred eleven UK residents aged 40 to 70 years recruited between 2006 and 2010 underwent retinal imaging.

Methods

Macular-centered OCT images acquired at the baseline visit were segmented for retinal sublayer thicknesses. Very severe periodontitis was ascertained through a touchscreen questionnaire. Linear mixed effects regression modeled the association between very severe periodontitis and retinal sublayer thicknesses, adjusting for age, sex, ethnicity, socioeconomic status, alcohol consumption, smoking status, diabetes mellitus, hypertension, refractive error, and previous cataract surgery.

Main Outcome Measures

Photoreceptor layer (PRL) and retinal pigment epithelium-Bruch's membrane (RPE-BM) thicknesses.

Results

Among 36 897 participants included in the analysis, 1571 (4.3%) reported very severe periodontitis. Affected individuals were older, lived in areas of greater socioeconomic deprivation, and were more likely to be hypertensive, diabetic, and current smokers (all P < 0.001). On average, those with very severe periodontitis were hyperopic (0.05 ± 2.27 diopters) while those unaffected were myopic (-0.29 ± 2.40 diopters, P < 0.001). Following adjusted analysis, very severe periodontitis was associated with thinner PRL (-0.55 μm, 95% confidence interval [CI], -0.97 to -0.12; P = 0.022) but there was no difference in RPE-BM thickness (0.00 μm, 95% CI, -0.12 to 0.13; P = 0.97). The association between PRL thickness and very severe periodontitis was modified by age (P < 0.001). Stratifying individuals by age, thinner PRL was seen among those aged 60 to 69 years with disease (-1.19 μm, 95% CI, -1.85 to -0.53; P < 0.001) but not among those aged < 60 years.

Conclusions

Among those with no known eye disease, very severe periodontitis is statistically associated with a thinner PRL, consistent with incipient AMD. Optimizing oral hygiene may hold additional relevance for people at risk of degenerative retinal disease.

Financial Disclosures

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

Can objective parameters in optical coherence tomography be useful markers in the treatment and follow-up of type 1 and type 2 macular neovascularizations related to neovascular age-related macular degeneration?

PURPOSE

The aim of this study was to compare the responses of type 1 and type 2 macular neovascularizations (MNV) caused by neovascular type age-related macular degeneration (n-AMD) to intravitreal anti-vascular endothelial growth factor (VEGF) treatments using quantitative parameters determined by optical coherence tomography (OCT). Additionally, it was also intended to assess the connections between these quantitative parameters and changes in best-corrected visual acuity (BCVA) and the number of intravitreal anti-VEGF injections required within a year.

MATERIALS AND METHODS

In our retrospective and observational study, the data of 90 eyes of 90 patients diagnosed with n-AMD and treated with intravitreal anti-VEGF with the "Pro re nata" method were evaluated. Subtypes of existing MNVs were distinguished with previously taken optical coherence tomography angiography (OCTA) images. In spectral domain OCT examinations, central macular thickness (CMT) and central macular volume (CMV) values were recorded at baseline and 12th month. The number of intravitreal anti-VEGF injections during the 12 month follow-up period was also recorded for each patient. Obtained data were compared between MNV types.

RESULTS

Of the n-AMD cases examined in the study, 56.66% had type 1 MNV and 43.34% had type 2 MNV. The mean baseline BCVA logMAR values in eyes with type 2 MNV (1.15 ± 0.43) were higher than those observed in eyes with type 1 MNV (0.76 ± 0.42) (p = 0.001). Similarly, mean baseline CMT and CMV values in eyes with type 2 MNV were higher than those observed in eyes with type 1 MNV (respectively 424.89 ± 49.46 μm vs. 341.39 ± 37.06 μm; 9.17 ± 0.89 μm3 vs. 8.49 ± 0.53 μm3; p < 0.05). After 12 months of treatment, logMAR values of BCVA (0.86 ± 0.42) in subjects with type 2 MNV were higher than those in subjects with type 1 MNV (0.57 ± 0.37) (p = 0.001). Mean CMT and CMV values at 12th month in subjects with type 2 MNV (379.11 ± 46.36 μm and 8.66 ± 0.79 μm3, respectively) were observed to be higher than those with type 1 MNV (296.95 ± 33.96 μm and 8.01 ± 0.52 mm3, respectively) (p < 0.05). In type 2 MNVs, positive correlations were observed between both baseline and 12th month BCVA logMAR values and baseline CMV (p < 0.05). Similarly, in type 2 MNVs, a positive correlation was observed between 12th month BCVA logMAR values and 12th month CMV (p < 0.05). The total number of intravitreal anti-VEGF injections at 12 months was similar in both groups (p = 0.851).

CONCLUSION

In this study, in which we performed a subtype analysis of MNV cases, we observed that the visual function was worse at the beginning and the end of the 12th month, and the CMT and CMV values were higher in the type 2 MNV group compared to the type 1 MNV cases. In addition, we found significant correlations between BCVA logMAR values and CMV values in type 2 MNV cases. In the follow-up of these cases, CMT, which is a more widely used quantitative method, and CMV, which is a newer OCT measurement parameter, may be more useful in patient follow-up and evaluation of treatment efficacy, especially for type 2 MNV cases.

Brolucizumab for recalcitrant macular neovascularization in age-related macular degeneration with pigment epithelial detachment

PURPOSE

To analyze anatomic and functional response to intravitreal brolucizumab in age-related macular degeneration recalcitrant to previous intravitreal anti-VEGF therapies.

METHODS

In this monocentric, one arm, retrospective study, eyes affected by neovascular age-related macular degeneration (nAMD) resistant to other intravitreally injected anti-vascular endothelial growth factor inhibitors were switched to intravitreal brolucizumab. All patients underwent ophthalmological examinations at baseline and in regular follow-up intervals. Best registered visual acuity (BRVA), Goldmann tonometry, intraocular pressure (IOP), central retinal thickness (CRT) and pigment epithelial detachment (PED) characteristics were analyzed at initiation of anti-VEGF treatment, at treatment switch, and at the end of brolucizumab loading phase.

RESULTS

The study included 20 eyes of 18 consecutively treated patients (age: 77 ± 6 years). All eyes had macular neovascularization with PED. Previous treatments included intravitreal aflibercept, bevacizumab, and ranibizumab and had not resulted in a significant improvement in BRVA (0.5 ± 0.5 logMAR vs 0.5 ± 0.6 logMAR) or mean CRT (320 ± 60 µm vs 313 ± 83 µm) up to treatment switch to brolucizumab. At the end of the brolucizumab loading phase, there was significant improvement for both BRVA (0.3 ± 0.2 logMAR, P < 0.05) and CRT (264 ± 55 µm, P < 0.05). Under previous anti-VEGF therapy, there was a significant increase/deterioration in both PED area (2.68 mm2 to 5.18 mm2, P < 0.05) and PED volume (0.39 mm3 to 1.07 mm3, P < 0.05); however, both parameters improved after switching to brolucizumab (3.81 mm2 and 0.37 mm3, P < 0.05).

CONCLUSION

Our results suggest a favourable anatomical and visual response after treatment switch to brolucizumab in patients with nAMD refractory to previous anti-VEGF agents.

Optical Coherence Tomography Features of Macular Hyperpigmented Lesions without Intraretinal Hyperreflective Foci in Age-Related Macular Degeneration

PURPOSE

To evaluate the optical coherence tomography (OCT) features of hyperpigmented lesions in the absence of intraretinal hyperreflective foci (IHRF) on OCT in eyes with age-related macular degeneration (AMD).

METHODS

We retrospectively analyzed OCT images of eyes with intermediate AMD (iAMD) and macular hyperpigmentation (HP) on color fundus photograph (CFP) but without IHRF on OCT in the corresponding location. The most prominent or definite HP was selected for analysis. The infrared reflectance (IR) image registered with the CFP, and the location corresponding to the HP lesion were defined on the IR image. The location of the HP on the corresponding OCT B-scan was assessed for retinal pigment epithelium (RPE) elevation, acquired vitelliform lesion (AVL), abnormal retinal pigment epithelium + basal lamina (RPE + BL) band reflectivity, RPE + BL band thickening, as well as interdigitation zone (IZ), ellipsoid zone (EZ) and external limiting membrane (ELM) disruption.

RESULTS

49 eyes (39 patients) were included in this study. Forty-six (94%) of the hyperpigmented lesions showed a thickened RPE + BL band. RPE + BL band reflectivity was increased in 37 (76%) of the lesions. RPE + BL band thickening, however, was not correlated with RPE + BL band reflectivity (p-value = 0.31). Either thickening or hyperreflectivity of the RPE + BL band was present in all cases. Twenty (41%) lesions had evidence of ELM disruption, 42 (86%) demonstrated EZ disruption and 48 (98%) had IZ disruption. Five (10%) HPs demonstrated AVL. Among cases with RPE elevation (15 cases, 31%), 10 were classified as drusen, 2 as drusenoid PEDs, and 3 as fibrovascular PEDs.

CONCLUSIONS

Thickening and/or hyperreflectivity of the RPE + BL band commonly correspond to regions of macular hyperpigmentation without IHRF in eyes with iAMD.

Explainable artificial intelligence model for the detection of geographic atrophy using colour retinal photographs

OBJECTIVE

To develop and validate an explainable artificial intelligence (AI) model for detecting geographic atrophy (GA) via colour retinal photographs.

METHODS AND ANALYSIS

We conducted a prospective study where colour fundus images were collected from healthy individuals and patients with retinal diseases using an automated imaging system. All images were categorised into three classes: healthy, GA and other retinal diseases, by two experienced retinologists. Simultaneously, an explainable learning model using class activation mapping techniques categorised each image into one of the three classes. The AI system's performance was then compared with manual evaluations.

RESULTS

A total of 540 colour retinal photographs were collected. Data was divided such that 300 images from each class trained the AI model, 120 for validation and 120 for performance testing. In distinguishing between GA and healthy eyes, the model demonstrated a sensitivity of 100%, specificity of 97.5% and an overall diagnostic accuracy of 98.4%. Performance metrics like area under the receiver operating characteristic (AUC-ROC, 0.988) and the precision-recall (AUC-PR, 0.952) curves reinforced the model's robust achievement. When differentiating GA from other retinal conditions, the model preserved a diagnostic accuracy of 96.8%, a precision of 90.9% and a recall of 100%, leading to an F1-score of 0.952. The AUC-ROC and AUC-PR scores were 0.975 and 0.909, respectively.

CONCLUSIONS

Our explainable AI model exhibits excellent performance in detecting GA using colour retinal images. With its high sensitivity, specificity and overall diagnostic accuracy, the AI model stands as a powerful tool for the automated diagnosis of GA.

Automated deep learning-based AMD detection and staging in real-world OCT datasets (PINNACLE study report 5)

Real-world retinal optical coherence tomography (OCT) scans are available in abundance in primary and secondary eye care centres. They contain a wealth of information to be analyzed in retrospective studies. The associated electronic health records alone are often not enough to generate a high-quality dataset for clinical, statistical, and machine learning analysis. We have developed a deep learning-based age-related macular degeneration (AMD) stage classifier, to efficiently identify the first onset of early/intermediate (iAMD), atrophic (GA), and neovascular (nAMD) stage of AMD in retrospective data. We trained a two-stage convolutional neural network to classify macula-centered 3D volumes from Topcon OCT images into 4 classes: Normal, iAMD, GA and nAMD. In the first stage, a 2D ResNet50 is trained to identify the disease categories on the individual OCT B-scans while in the second stage, four smaller models (ResNets) use the concatenated B-scan-wise output from the first stage to classify the entire OCT volume. Classification uncertainty estimates are generated with Monte-Carlo dropout at inference time. The model was trained on a real-world OCT dataset, 3765 scans of 1849 eyes, and extensively evaluated, where it reached an average ROC-AUC of 0.94 in a real-world test set.

Hyperreflective Material Boundary Remodeling in Neovascular Age-Related Macular Degeneration: A Post Hoc Analysis of the AVENUE Trial

OBJECTIVE

To describe the spatial and temporal characteristics of hyperreflective material (HRM) on spectral-domain OCT (SD-OCT) in neovascular age-related macular degeneration (nAMD) during antiangiogenic treatment and explore associations with best-corrected visual acuity (BCVA) and macular atrophy (MA).

DESIGN

Retrospective regrading of SD-OCT-images from the multicenter, randomized controlled AVENUE trial (NCT02484690, conducted from August 2015 to September 2017).

PARTICIPANTS

Treatment-naive nAMD patients enrolled from 50 sites in the US.

METHODS

Retrospective regrading and secondary analysis.

MAIN OUTCOME MEASURES

Spectral-domain OCT images from 207 study eyes that fit criteria for the present analysis were graded for HRM features, its evolution, and associated hypertransmission into choroid (HTC), a proxy for MA. The appearance of a well-defined hyperreflective inner boundary that separated persistent HRM from the neurosensory retina continuous with the adjacent retinal pigment epithelium layer was defined as hyperreflective material boundary remodeling (HRM-BR). Patterns of HRM composition/evolution were defined as follows: (1) no subretinal HRM at baseline, (2) fully resolved, (3) persistent with complete HRM-BR, or (4) partial/absent HRM-BR. Associations of HRM patterns with BCVA and HTC were analyzed. Predictive factors for complete HRM-BR were explored.

RESULTS

Of 207 included eyes, subretinal HRM was present in 159 (76.8%) at baseline and persisted until month 9 in 118 (57.0%) eyes. Of these 118 eyes, 44.9% developed complete HRM-BR and had similar BCVA outcomes by month 9 compared with no/fully resolved subretinal HRM. Partial/absent HRM-BR had a strong negative association with BCVA outcome (-6.1 ETDRS letters; P = 0.016) and a higher frequency of intralesional HTC (69.2%) compared with eyes with complete HRM-BR (20.8%) at month 9. Older age (odds ratio [OR], 0.96; P = 0.054) and presence of intralesional HTC (OR, 0.06; P = 0.010) at baseline were associated with lower odds of complete HRM-BR at month 9.

CONCLUSIONS

In nAMD eyes under antiangiogenic treatment, complete HRM-BR occurred frequently and was associated with better BCVA than when HRM-BR was only partial/absent.

FINANCIAL DISCLOSURE(S)

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

Few-shot out-of-distribution detection for automated screening in retinal OCT images using deep learning

Deep neural networks have been increasingly proposed for automated screening and diagnosis of retinal diseases from optical coherence tomography (OCT), but often provide high-confidence predictions on out-of-distribution (OOD) cases, compromising their clinical usage. With this in mind, we performed an in-depth comparative analysis of the state-of-the-art uncertainty estimation methods for OOD detection in retinal OCT imaging. The analysis was performed within the use-case of automated screening and staging of age-related macular degeneration (AMD), one of the leading causes of blindness worldwide, where we achieved a macro-average area under the curve (AUC) of 0.981 for AMD classification. We focus on a few-shot Outlier Exposure (OE) method and the detection of near-OOD cases that share pathomorphological characteristics with the inlier AMD classes. Scoring the OOD case based on the Cosine distance in the feature space from the penultimate network layer proved to be a robust approach for OOD detection, especially in combination with the OE. Using Cosine distance and only 8 outliers exposed per class, we were able to improve the near-OOD detection performance of the OE with Reject Bucket method by [Formula: see text] 10% compared to without OE, reaching an AUC of 0.937. The Cosine distance served as a robust metric for OOD detection of both known and unknown classes and should thus be considered as an alternative to the reject bucket class probability in OE approaches, especially in the few-shot scenario. The inclusion of these methodologies did not come at the expense of classification performance, and can substantially improve the reliability and trustworthiness of the resulting deep learning-based diagnostic systems in the context of retinal OCT.

Fibrosis in neovascular age-related macular degeneration: A review of definitions based on clinical imaging

Despite the success of antiangiogenic therapy in controlling exudation in neovascular age-related macular degeneration (nAMD), the involvement of the outer retina in fibrosis results in gradual vision loss over time. The development of drugs that prevent or ameliorate fibrosis in nAMD requires that it is accurately detected and quantified with reliable endpoints and identification of robust biomarkers. Achievement of such an aim is currently challenging due to the lack of a consensus definition of fibrosis in nAMD. As a first step towards the establishment of a clear definition of fibrosis, we provide an extensive overview of the imaging modalities and criteria used to characterize fibrosis in nAMD. We observed variety in the selection of individual and combinations of imaging modalities, and criteria for detection. We also observed heterogeneity in classification systems and severity scales for fibrosis. The most commonly used imaging modalities were color fundus photography, fluorescein angiography and optical coherence tomography (OCT). A multimodal approach was frequently utilized. Our review suggests that OCT offers a more detailed, objective and sensitive characterization than color fundus photography/fluorescein angiography. Thus, we recommend it as a primary modality for fibrosis evaluation. This review provides a basis for future discussions to reach a consensus definition using standardized terms based on a detailed characterization of fibrosis, its presence and evolution, and taking into consideration impact on visual function. Achieving this goal is of paramount importance for the development of antifibrotic therapies.

Optical coherence tomography imaging biomarkers associated with neovascular age-related macular degeneration: a systematic review

The aim of this systematic literature review is twofold, (1) detail the impact of retinal biomarkers identifiable via optical coherence tomography (OCT) on disease progression and response to treatment in neovascular age-related macular degeneration (nAMD) and (2) establish which biomarkers are currently identifiable by artificial intelligence (AI) models and the utilisation of this technology. Following the PRISMA guidelines, PubMed was searched for peer-reviewed publications dated between January 2016 and January 2022.

POPULATION

Patients diagnosed with nAMD with OCT imaging.

SETTINGS

Comparable settings to NHS hospitals.

STUDY DESIGNS

Randomised controlled trials, prospective/retrospective cohort studies and review articles. From 228 articles, 130 were full-text reviewed, 50 were removed for falling outside the scope of this review with 10 added from the author's inventory, resulting in the inclusion of 90 articles. From 9 biomarkers identified; intraretinal fluid (IRF), subretinal fluid, pigment epithelial detachment, subretinal hyperreflective material (SHRM), retinal pigmental epithelial (RPE) atrophy, drusen, outer retinal tabulation (ORT), hyperreflective foci (HF) and retinal thickness, 5 are considered pertinent to nAMD disease progression; IRF, SHRM, drusen, ORT and HF. A number of these biomarkers can be classified using current AI models. Significant retinal biomarkers pertinent to disease activity and progression in nAMD are identifiable via OCT; IRF being the most important in terms of the significant impact on visual outcome. Incorporating AI into ophthalmology practice is a promising advancement towards automated and reproducible analyses of OCT data with the ability to diagnose disease and predict future disease conversion.

SYSTEMATIC REVIEW REGISTRATION

This review has been registered with PROSPERO (registration ID: CRD42021233200).

Second eyes to develop neovascular age-related macular degeneration have fewer symptoms and better one-year visual outcomes

BACKGROUND

This study compares the visual and anatomical outcomes for the eyes of patients who developed sequential neovascular age-related macular degeneration (nAMD), both at the time of diagnosis and at one year after treatment.

METHODS

The study comprised a retrospective case series of 52 patients whose eyes were diagnosed sequentially with nAMD. All eyes were treated with three monthly loading doses of anti-vascular endothelial growth factor agents, followed by further intravitreal injections, as required. Baseline characteristics and outcomes at one year after diagnosis and initial treatment were compared between first and second eyes and included visual acuity (VA), central macular thickness (CMT), and pigment epithelial detachment (PED) height on optical coherence tomography (OCT) imaging.

RESULTS

VA at diagnosis was better for second eyes compared with first eyes to develop nAMD (logMAR 0.68 ± 0.51 versus logMAR 0.41 ± 0.34, P = 0.002) and remained so at one year (logMAR 0.61 ± 0.60 versus logMAR 0.42 ± 0.37, P = 0.041). Similarly, PED height at diagnosis was higher in first eyes (225 ± 176 μm versus 155 ± 144 μm, P = 0.003) and also at one year (188 ± 137 μm versus 140 ± 112 μm, P = 0.019). Whereas most patients reported symptoms at first eye diagnosis (71.2%), half as many second eyes were symptomatic (28.8%, P < 0.001). Significantly more symptomatic first eyes experienced visual distortions (32.4% versus 13.3%) or scotomas (29.4% versus 6.7%), compared with a less specific visual complaint of blurry vision (38.2% versus 80.0%, P = 0.006).

CONCLUSIONS

Compared with first eyes to develop nAMD, second eyes tended to have better vision, smaller PED heights, and fewer symptoms likely because monitoring permitted earlier diagnosis.

Optometrists' attitudes toward using OCT angiography lag behind other retinal imaging types

PURPOSE

While optometrists' attitudes toward established retinal imaging types are generally positive, they are unknown for optical coherence tomography angiography (OCTA). We performed a cross-sectional survey to estimate attitudes toward OCTA and identify clinician and/or practice characteristics that influence them.

METHODS

A paper-based survey was mailed to 252 randomly selected optometrists in Australia. Five-point Likert-scale items from a previous survey assessing attitudes toward new technology were included to probe respondent characteristics and attitudes toward retinal imaging. Performance expectancy attitudes toward OCTA were elicited by the statement 'I believe OCTA is useful in daily practice'. Mean scores out of five (mean [SD]) were rounded and mapped to appropriate descriptive statements.

RESULTS

The response rate was 47% (118/252). The mean (SD) age of respondents was 44.0 (13.8) years and 50.8% (60/118) were female. Optometrists had 19.9 (14.0) years of clinical experience and 66.9% (79/118) worked at independent practices. In total, 8.5% (10/118) of respondents used OCTA to provide clinical care. Optometrists agreed that optical coherence tomography (OCT), colour fundus imaging, ultra-wide field imaging and fundus autofluorescence (mean scores 3.6-4.7 out of 5) were useful in daily practice but felt neutral about whether OCTA was useful (3.4 [0.8]). Optometrists believed that OCTA was less enjoyable to use (p < 0.0001), less endorsed by peers (p < 0.0001) and felt less confident that they had the knowledge to interpret OCTA (p < 0.0001) compared to other retinal imaging types.

CONCLUSIONS

Optometrists are undecided on whether OCTA is useful in daily practice and had lower expectations that using OCTA would confer job performance benefits compared to other retinal imaging types. Further work is needed to advocate the benefits of using OCTA across the profession.

Age-related macular degeneration

Age-related macular degeneration is an increasingly important public health issue due to ageing populations and increased longevity. Age-related macular degeneration affects individuals older than 55 years and threatens high-acuity central vision required for important tasks such as reading, driving, and recognising faces. Advances in retinal imaging have identified biomarkers of progression to late age-related macular degeneration. New treatments for neovascular age-related macular degeneration offer potentially longer-lasting effects, and progress is being made towards a treatment for atrophic late age-related macular degeneration. An effective intervention to slow progression in the earlier stages of disease, or to prevent late age-related macular degeneration development remains elusive, and our understanding of underlying mechanistic pathways continues to evolve.

Retromode Imaging in Age-Related Macular Degeneration

Background and Objectives: Retromode is a relatively new retinal-imaging technique that is based on the transillumination principle and is obtained with a scanning laser ophthalmoscope that uses light in the infrared spectrum. The laser light penetrates into the deep retinal layers and the choroid. Retromode images are captured with a laterally displaced aperture, and the detector captures only the scattered light. The result is a high-contrast pseudo-three-dimensional image. Age-related macular degeneration (AMD) is a disabling retinal disease. AMD is characterized in its early stage by small and intermediate drusen formation, while the signs of intermediate AMD are large drusen and/or pigmentary abnormalities. Late AMD has two forms, geographic atrophy, which is the advanced form of dry AMD, and wet AMD. Most of the lesions of AMD are located in the outer layers of the retina. This new imaging method can provide a glimpse of the deep retinal layers' topographic changes in a non-invasive, fast, and effective way that can match the other imaging tools available. Materials and Methods: The literature review was performed by searching the PubMed database using the following combination of keywords: retromode imaging and age-related macular degeneration. Relevant images similar to the ones in the literature were identified and used as models. Results: The purpose of this article is to highlight the utility of incorporating retromode imaging into the multimodal evaluation of the retina in patients with AMD and to gather and integrate these findings into a brief but comprehensive paper. Conclusions: Retromode imaging is a good screening, diagnosis, and monitoring tool for patients with AMD.

An Explainable Fully Dense Fusion Neural Network with Deep Support Vector Machine for Retinal Disease Determination

Retinal issues are crucial because they result in visual loss. Early diagnosis can aid physicians in initiating treatment and preventing visual loss. Optical coherence tomography (OCT), which portrays retinal morphology cross-sectionally and noninvasively, is used to identify retinal abnormalities. The process of analyzing OCT images, on the other hand, takes time. This study has proposed a hybrid approach based on a fully dense fusion neural network (FD-CNN) and dual preprocessing to identify retinal diseases, such as choroidal neovascularization, diabetic macular edema, drusen from OCT images. A dual preprocessing methodology, in other words, a hybrid speckle reduction filter was initially used to diminish speckle noise present in OCT images. Secondly, the FD-CNN architecture was trained, and the features obtained from this architecture were extracted. Then Deep Support Vector Machine (D-SVM) and Deep K-Nearest Neighbor (D-KNN) classifiers were proposed to reclassify those features and tested on University of California San Diego (UCSD) and Duke OCT datasets. D-SVM demonstrated the best performance in both datasets. D-SVM achieved 99.60% accuracy, 99.60% sensitivity, 99.87% specificity, 99.60% precision and 99.60% F1 score in the UCSD dataset. It achieved 97.50% accuracy, 97.64% sensitivity, 98.91% specificity, 96.61% precision, and 97.03% F1 score in Duke dataset. Additionally, the results were compared to state-of-the-art works on the both datasets. The D-SVM was demonstrated to be an efficient and productive strategy for improving the robustness of automatic retinal disease classification. Also, in this study, it is shown that the unboxing of how AI systems' black-box choices is made by generating heat maps using the local interpretable model-agnostic explanation method, which is an explainable artificial intelligence (XAI) technique. Heat maps, in particular, may contribute to the development of more stable deep learning-based systems, as well as enhancing the confidence in the diagnosis of retinal disease in the analysis of OCT image for ophthalmologists.

Molecular mechanisms of TGFβ-mediated EMT of retinal pigment epithelium in subretinal fibrosis of age-related macular degeneration

Age-related macular degeneration (AMD) is one of the leading causes of blindness in the elderly, affecting the macula of the retina and resulting in vision loss. There are two types of AMD, wet and dry, both of which cause visual impairment. Wet AMD is called neovascular AMD (nAMD) and is characterized by the formation of choroidal neovascular vessels (CNVs) in the macula. nAMD can be treated with intravitreal injections of vascular endothelial growth factor (VEGF) inhibitors, which help improve vision. However, approximately half the patients do not achieve satisfactory results. Subretinal fibrosis often develops late in nAMD, leading to irreversible photoreceptor degeneration and contributing to visual loss. Currently, no treatment exists for subretinal fibrosis, and the molecular mechanisms of fibrous tissue formation following neovascular lesions remain unclear. In this review, we describe the clinical features and molecular mechanisms of macular fibrosis secondary to nAMD. Myofibroblasts play an essential role in the development of fibrosis. This review summarizes the latest findings on the clinical features and cellular and molecular mechanisms of the pathogenesis of subretinal fibrosis in nAMD and discusses the potential therapeutic strategies to control subretinal fibrosis in the future.

Consequences of Real-World Surveillance of Fellow Eyes in Neovascular Age-Related Macular Degeneration

This study investigated whether the interval of monitoring at-risk, fellow eyes of patients with unilateral neovascular age-related macular degeneration (nAMD) has any bearing on the severity of the disease at the time of diagnosis. The study comprised a retrospective, cross-sectional comparative case series of treatment-naïve eyes in patients who were diagnosed sequentially with nAMD. We compared the visual acuity (VA) and central macular thickness (CMT) of patients who were actively receiving intravitreal injections (IVIs) of anti-vascular endothelial growth factor (anti-VEGF) agents at the time of second eye diagnosis with the VA and CMT of patients who had ceased treatment in their first eye because of reaching end-stages of disease. Intervals of visits and frequency of monitoring the macula of fellow eyes by means of optical coherence tomography (OCT) were abstracted from the medical record. We found that the at-risk fellow eyes of patients who had stopped treatment for nAMD in their first eye prior to fellow eye conversion were monitored significantly less frequently than the fellow eyes of patients who continued to receive treatment at the time of second eye diagnosis. Despite less frequent monitoring, VA and CMT were similar at the time of fellow eye diagnosis for both groups.

Hyper-Reflective Foci in Intermediate Age-Related Macular Degeneration: Spatial Abundance and Impact on Retinal Morphology

Purpose

The purpose of this study was to analyze spatially resolved structural changes at retinal locations in presence (+) or absence (-) of hyper-reflective foci (HRF) in eyes with subretinal pigment epithelium (RPE) drusen in intermediate age-related macular degeneration (iAMD).

Methods

Patients with IAMD (n = 40; mean age = 69.7 ± 9.2 [SD] years) and healthy controls (n = 27; 64.2 ± 9.0) underwent spectral-domain optical-coherence-tomography imaging and fundus-controlled perimetry testing. After reviewing retinal layer segmentation, presence of HRF was annotated and retinal layer thicknesses (RLTs) extracted using ImageJ. Localized RLTs were compared between +HRF and -HRF positions. Univariate mixed linear models were used to investigate associations among RLT, HRF presence, and HRF size.

Results

In iAMD eyes, a mean of 11.1 ± 12.5 HRF were detected with a peak abundance at 0.5 to 1.5 mm eccentricity to the fovea. At +HRF positions, outer nuclear layer (ONL; P = 0.0013, average difference = -12.4 µm) and retinal pigment epithelium drusen complex (RPEDC; P < 0.0001, +45.6 µm) thicknesses differed significantly compared to -HRF positions, even after correcting for accompanying drusen-related RPEDC layer thickening (P = 0.01). Mixed linear models revealed a significant association between increasing HRF area and decreasing ONL (association score = -0.17, P < 0.0001; 95% confidence interval [CI] = -0.22 to -0.11), and inner photoreceptor segments (IS) layer thicknesses (-0.08, P = 0.005; 95% CI = -0.14 to -0.03). Spearman rank correlation analysis yielded a significant correlation between total HRF area and mesopic (P = 0.015), but not scotopic (P = 0.305) retinal sensitivity losses.

Conclusions

Descriptive analysis of this study demonstrated a predominant distribution of HRF at a foveal eccentricity of 0.5 to 1.5 mm, whereas further refined topographic analysis revealed a significant ONL layer thinning in presence of HRF even after correction for sub-RPE drusen presence compared to lesions in absence of HRF. Longitudinal studies are further needed to analyze the prognostic impact as well as the role of HRF presence in the context of iAMD.

Pigment epithelial detachment composition indices (PEDCI) in neovascular age-related macular degeneration

We provide an automated analysis of the pigment epithelial detachments (PEDs) in neovascular age-related macular degeneration (nAMD) and estimate areas of serous, neovascular, and fibrous tissues within PEDs. A retrospective analysis of high-definition spectral-domain OCT B-scans from 43 eyes of 37 patients with nAMD with presence of fibrovascular PED was done. PEDs were manually segmented and then filtered using 2D kernels to classify pixels within the PED as serous, neovascular, or fibrous. A set of PED composition indices were calculated on a per-image basis using relative PED area of serous (PEDCI-S), neovascular (PEDCI-N), and fibrous (PEDCI-F) tissue. Accuracy of segmentation and classification within the PED were graded in masked fashion. Mean overall intra-observer repeatability and inter-observer reproducibility were 0.86 ± 0.07 and 0.86 ± 0.03 respectively using intraclass correlations. The mean graded scores were 96.99 ± 8.18, 92.12 ± 7.97, 91.48 ± 8.93, and 92.29 ± 8.97 for segmentation, serous, neovascular, and fibrous respectively. Mean (range) PEDCI-S, PEDCI-N, and PEDCI-F were 0.253 (0-0.952), 0.554 (0-1), and 0.193 (0-0.693). A kernel-based image processing approach demonstrates potential for approximating PED composition. Evaluating follow up changes during nAMD treatment with respect to PEDCI would be useful for further clinical applications.

Learning Asymmetric and Local Features in Multi-Dimensional Data Through Wavelets With Recursive Partitioning

Effective learning of asymmetric and local features in images and other data observed on multi-dimensional grids is a challenging objective critical for a wide range of image processing applications involving biomedical and natural images. It requires methods that are sensitive to local details while fast enough to handle massive numbers of images of ever increasing sizes. We introduce a probabilistic model-based framework that achieves these objectives by incorporating adaptivity into discrete wavelet transforms (DWT) through Bayesian hierarchical modeling, thereby allowing wavelet bases to adapt to the geometric structure of the data while maintaining the high computational scalability of wavelet methods-linear in the sample size (e.g., the resolution of an image). We derive a recursive representation of the Bayesian posterior model which leads to an exact message passing algorithm to complete learning and inference. While our framework is applicable to a range of problems including multi-dimensional signal processing, compression, and structural learning, we illustrate its work and evaluate its performance in the context of image reconstruction using real images from the ImageNet database, two widely used benchmark datasets, and a dataset from retinal optical coherence tomography and compare its performance to state-of-the-art methods based on basis transforms and deep learning.

Extending field-of-view of retinal imaging by optical coherence tomography using convolutional Lissajous and slow scan patterns

Optical coherence tomography (OCT) is a high-speed non-invasive cross-sectional imaging technique. Although its imaging speed is high, three-dimensional high-spatial-sampling-density imaging of in vivo tissues with a wide field-of-view (FOV) is challenging. We employed convolved Lissajous and slow circular scanning patterns to extend the FOV of retinal OCT imaging with a 1-µm, 100-kHz-sweep-rate swept-source OCT prototype system. Displacements of sampling points due to eye movements are corrected by post-processing based on a Lissajous scan. Wide FOV three-dimensional retinal imaging with high sampling density and motion correction is achieved. Three-dimensional structures obtained using repeated imaging sessions of a healthy volunteer and two patients showed good agreement. The demonstrated technique will extend the FOV of simple point-scanning OCT, such as commercial ophthalmic OCT devices, without sacrificing sampling density.

Etiology of Macular Edema Defined by Deep Learning in Optical Coherence Tomography Scans

Purpose

To develop an automated method based on deep learning (DL) to classify macular edema (ME) from the evaluation of optical coherence tomography (OCT) scans.

Methods

A total of 4230 images were obtained from data repositories of patients attended in an ophthalmology clinic in Colombia and two free open-access databases. They were annotated with four biomarkers (BMs) as intraretinal fluid, subretinal fluid, hyperreflective foci/tissue, and drusen. Then the scans were labeled as control or ocular disease among diabetic macular edema (DME), neovascular age-related macular degeneration (nAMD), and retinal vein occlusion (RVO) by two expert ophthalmologists. Our method was developed by following four consecutive phases: segmentation of BMs, the combination of BMs, feature extraction with convolutional neural networks to achieve binary classification for each disease, and, finally, multiclass classification of diseases and control images.

Results

The accuracy of our model for nAMD was 97%, and for DME, RVO, and control were 94%, 93%, and 93%, respectively. Area under curve values were 0.99, 0.98, 0.96, and 0.97, respectively. The mean Cohen's kappa coefficient for the multiclass classification task was 0.84.

Conclusions

The proposed DL model may identify OCT scans as normal and ME. In addition, it may classify its cause among three major exudative retinal diseases with high accuracy and reliability.

Translational Relevance

Our DL approach can optimize the efficiency and timeliness of appropriate etiological diagnosis of ME, thus improving patient access and clinical decision making. It could be useful in places with a shortage of specialists and for readers that evaluate OCT scans remotely.

DHNet: High-resolution and hierarchical network for cross-domain oct speckle noise reduction

PURPOSE

Optical coherence tomography (OCT) imaging uses the principle of Michelson interferometry to obtain high-resolution images by coherent superposing of multiple forward and backward scattered light waves with random phases. This process inevitably produces speckle noise that severely compromises visual quality of OCT images and degrades performances of subsequent image analysis tasks. In addition, datasets obtained by different OCT scanners have distribution shifts, making a speckle noise suppression model difficult to be generalized across multiple datasets. In order to solve the above issues, we propose a novel end-to-end denoising framework for OCT images collected by different scanners.

METHODS

The proposed model utilizes the high-resolution network (HRNet) as backbone for image restoration, which reconstructs high fidelity images by maintaining high-resolution representations throughout the entire learning process. To compensate distribution shifts among datasets collected by different scanners, we develop a hierarchical adversarial learning strategy for domain adaption. The proposed model is trained using datasets with clean ground truth produced by two commercial OCT scanners, and then applied to suppress speckle noise in OCT images collected by our recently developed OCT scanner, BV-1000 (China Bigvision Corporation). We name the proposed model as DHNet (Double-H-Net, High-resolution and Hierarchical Network).

RESULTS

We compare DHNet with state-of-the-art methods and experiment results show that DHNet improves signal-to-noise ratio (SNR) by a large margin of 18.137dB as compared to the best of our previous method. In addition, DHNet achieves a testing time of 25ms, which satisfies the real-time processing requirement for the BV-1000 scanner. We also conduct retinal layer segmentation experiment on OCT images before and after denoising and show that DHNet can also improve segmentation.

CONCLUSIONS

The proposed DHNet can compensate domain shifts between different datasets while significantly improve speckle noise suppression. The HRNet backbone is utilized to carry low- and high-resolution information to recover fidelity images. Domain adaptation is achieved by a hierarchical module through adversarial learning. In addition, DHNet achieved a testing time of 25ms, which satisfied the real-time processing requirement. This article is protected by copyright. All rights reserved.

Multispectral pattern recognition measures change in drusen area in age-related macular degeneration with high congruency to expert graders

Drusen are a hallmark lesion of age-related macular degeneration (AMD) and changes in their area and/or volume are strongly associated with disease progression. Assessment of longitudinal change in drusen size in clinical practice however is limited to a single commercial tool or manual inspection by clinicians. In this study we analysed change in drusen area in 33 eyes with intermediate AMD across two separate visits using a novel technique known as multispectral pattern recognition for en face retinal images from various imaging modalities (infrared (815 nm), fundus autofluorescence (488 nm) and green (532 nm) scanning laser ophthalmoscopy). We found 91% (30/33 eyes) agreement in the direction of drusen change for multispectral pattern recognition relative to expert graders who graded eyes as having drusen progression, regression or being stable. Multispectral pattern recognition showed 100% sensitivity (22/22 eyes) and 73% specificity (8/11 eyes). In comparison, we found only 70% (23/33 eyes) agreement in the direction of drusen change with a commercially available change analysis software, the Cirrus Advanced RPE Analysis relative to expert graders, with a sensitivity 64% (14/22 eyes) and specificity of 82% (9/11 eyes). Total drusen area or amount of change between visits had no significant effect on agreement. This suggests multispectral pattern recognition can quantify longitudinal change in drusen area from multimodal imaging with greater congruency to expert graders than a commercially available platform based on a single imaging modality. Considering the association of drusen area and disease progression, this method could aid clinical assessment and monitoring of AMD.

Masking of macular neovascular membranes by subretinal hyperreflective material on optical coherence tomography angiography

PURPOSE

The purpose of this study was to determine whether subretinal hyperreflective material (SHRM) may mask the detection of macular neovascular membranes (MNV) on optical coherence tomography angiography (OCTA).

METHODS

In this observational study, eyes with active neovascular age-related macular degeneration (nAMD), co-existing SHRM & intraretinal or subretinal fluid or hemorrhage on structural OCT, underwent OCTA & fundus fluorescein angiography (FFA) imaging. 6 × 6 mm choriocapillaris and outer retinal slabs on OCTA were examined to determine the presence of MNV underneath the SHRM. The corresponding area on FFA was used as a reference arm to confirm activity.

RESULTS

Thirty eyes of thirty patients with SHRM and active nAMD were recruited. All eyes failed to show a MNV in the choriocapillaris & avascular slabs of the OCTA underneath the SHRM, but showed active hyperfluorescent MNVs that increased in size and intensity in the late stages of FFA. In one eye, parts of a MNV under the SHRM were undetectable due to signal attenuation, while parts extending beyond the SHRM were detected on the choriocapillaris en face slab with flow on the B scan.

CONCLUSIONS

SHRM may act as a reflecting surface that limits the passage of light waves in OCTA, creating areas of signal attenuation and diminishing its ability to detect underlying MNVs.

Progression of Age-Related Macular Degeneration Among Individuals Homozygous for Risk Alleles on Chromosome 1 (CFH-CFHR5) or Chromosome 10 (ARMS2/HTRA1) or Both

IMPORTANCE

Age-related macular degeneration (AMD) is a common cause of irreversible vision loss among individuals older than 50 years. Although considerable advances have been made in our understanding of AMD genetics, the differential effects of major associated loci on disease manifestation and progression may not be well characterized.

OBJECTIVE

To elucidate the specific associations of the 2 most common genetic risk loci for AMD, the CFH-CFHR5 locus on chromosome 1q32 (Chr1) and the ARMS2/HTRA1 locus on chromosome 10q26 (Chr10)-independent of one another and in combination-with time to conversion to late-stage disease and to visual acuity loss.

DESIGN, SETTING, AND PARTICIPANTS

This case series study included 502 individuals who were homozygous for risk variants at both Chr1 and Chr10 (termed Chr1&10-risk) or at either Chr1 (Chr1-risk) or Chr10 (Chr10-risk) and who had enrolled in Genetic and Molecular Studies of Eye Diseases at the Sharon Eccles Steele Center for Translational Medicine between September 2009 and March 2020. Multimodal imaging data were reviewed for AMD staging, including grading of incomplete and complete retinal pigment epithelium and outer retinal atrophy.

MAIN OUTCOMES AND MEASURES

Hazard ratios and survival times for conversion to any late-stage AMD, atrophic or neovascular, and associated vision loss of 2 or more lines.

RESULTS

In total, 317 participants in the Chr1-risk group (median [IQR] age at first visit, 75.6 [69.5-81.7] years; 193 women [60.9%]), 93 participants in the Chr10-risk group (median [IQR] age at first visit, 77.5 [72.2-84.2] years; 62 women [66.7%]), and 92 participants in the Chr1&10-risk group (median [IQR] age at first visit, 71.7 [68.0-76.3] years; 62 women [67.4%]) were included in the analyses. After adjusting for age and AMD grade at first visit, compared with 257 participants in the Chr1-risk group, 56 participants in the Chr1&10-risk group (factor of 3.3 [95% CI, 1.6-6.8]; P < .001) and 58 participants in the Chr10-risk group (factor of 2.6 [95% CI, 1.3-5.2]; P = .007) were more likely to convert to a late-stage phenotype during follow-up. This difference was mostly associated with conversion to macular neovascularization, which occurred earlier in participants with Chr1&10-risk and Chr10-risk. Eyes in the Chr1&10-risk group (median [IQR] survival, 5.7 [2.1-11.1] years) were 2.1 (95% CI, 1.1-3.9; P = .03) times as likely and eyes in the Chr10-risk group (median [IQR] survival, 6.3 [2.7-11.3] years) were 1.8 (95% CI, 1.0-3.1; P = .05) times as likely to experience a visual acuity loss of 2 or more lines compared with eyes of the Chr1-risk group (median [IQR] survival, 9.4 [4.1-* (asterisk indicates event rate did not reach 75%)] years).

CONCLUSIONS AND RELEVANCE

These findings suggest differential associations of the 2 major AMD-related risk loci with structural and functional disease progression and suggest distinct underlying biological mechanisms associated with these 2 loci. These genotype-phenotype associations may warrant consideration when designing and interpreting AMD research studies and clinical trials.

Histology and Clinical Lifecycle of Acquired Vitelliform Lesion, a Pathway to Advanced Age-Related Macular Degeneration

PURPOSE

To evaluate hypotheses about the role of acquired vitelliform lesion (AVL) in age-related macular degeneration pathophysiology.

DESIGN

Laboratory histology study; retrospective, observational case series.

METHODS

Two donor eyes in a research archive with AVL and age-related macular degeneration were analyzed with light and electron microscopy for AVL content at locations matched to ex vivo B-scans. A retrospective, observational clinical cohort study of 42 eyes of 30 patients at 2 referral clinics determined the frequency of optical coherence tomography features stratified by AVL fate.

RESULTS

Histologic and clinical cases showed subretinal drusenoid deposit and drusen. Ultrastructural AVL components in 2 donor eyes included retinal pigment epithelium (RPE) organelles (3%-22% of volume), outer segments (2%-10%), lipid droplets (0.2%-12%), and a flocculent material (57%-59%). Of 48 AVLs (mean follow-up 46 ± 39 months), 50% collapsed to complete RPE and outer retinal atrophy, 38% were stable, 10% resorbed, and 2% developed neovascularization. The Early Treatment Diabetic Retinopathy Study grid central subfield contained 77% of AVLs. Hyperreflective foci, ellipsoid zone disruption, and hyperreflective thickening of the RPE-basal lamina-Bruch membrane band were common at maximum AVL expansion. Collapsing and noncollapsing AVLs had different growth rates (rapid vs slow, respectively).

CONCLUSIONS

AVL deposits contain unexpectedly low levels of RPE organelles and outer segments. Subfoveal predilection, reflectivity on optical coherence tomography, hyperautofluorescence, yellow color, and growth-regression phases suggest dysregulation of lipid transfer pathways specific to cone photoreceptors and supporting cells in formation of AVL deposit, analogous to drusen and subretinal drusenoid deposit. Prediction of AVL outcomes via growth rates should be confirmed in larger clinical studies.

Treatment Outcome of Wet Age-Related Macular Degeneration Management in Thailand: A Retrospective Real-World Study (TOWER Study)

INTRODUCTION

To present real-world outcomes of neovascular age-related macular degeneration (nAMD) management in Thailand.

METHODS

This multicenter retrospective study reviewed medical records of naive nAMD patients diagnosed from 1 January 2016 until 31 December 2018. The patients received at least one intravitreal anti-vascular endothelial growth factor (VEGF) treatment and had captured visual acuity (VA) at baseline and at month 12. Treatment outcomes were assessed at month 12, 24, and 36. The primary outcome was a mean change in VA from baseline to month 12.

RESULTS

Five hundred seventy-two (572) eyes were included in this study and of these eyes, 222 and 96 had 2- and 3-year follow-up periods, respectively. At month 12, the mean improvement of VA (ETDRS letter) was six letters (P < 0.0001), and central retinal thickness (CRT) decreased on average by 104 microns (P < 0.0001). However, visual improvement by 0.1 letters at month 36 did not show statistical significance. The presence of fluid was found in approximately half of patients throughout the study period (45.98%, 48.85%, and 50.91% at month 12, 24, and 36, respectively). Mean number of injections (SD) was 6.06 (3.00), 3.44 (2.94), and 2.71 (3.07) for years 1, 2, and 3, respectively. The mean number of visits (SD) in year 1 was 9.01 (2.60) and declined to 5.67 (2.69) in year 2 and 4.93 (2.49) in year 3. Patients who had an average injection interval of ≤ 8 weeks were 74.46% in year 1, 51.28% in year 2, and 45.24 in year 3; 35.31% of patients were lost to follow-up.

CONCLUSIONS

This analysis reflects real-world nAMD management with significant improvement of outcomes. At the same time, the study reveals unmet needs in anti-VEGF therapy in nAMD including persistent disease activities, inadequacy of available treatment, and lack of treatment adherence leading to visual deterioration in the long-term.

Osteopontin accumulates in basal deposits of human eyes with age-related macular degeneration and may serve as a biomarker of aging

A common clinical phenotype of several neurodegenerative and systemic disorders including Alzheimer's disease and atherosclerosis is the abnormal accumulation of extracellular material, which interferes with routine cellular functions. Similarly, patients with age-related macular degeneration (AMD), the leading cause of vision loss among the aged population, present with extracellular lipid- and protein-filled basal deposits in the back of the eye. While the exact mechanism of growth and formation of these deposits is poorly understood, much has been learned from investigating their composition, providing critical insights into AMD pathogenesis, prevention, and therapeutics. We identified human osteopontin (OPN), a phosphoprotein expressed in a variety of tissues in the body, as a newly discovered component of basal deposits in AMD patients, with a distinctive punctate staining pattern. OPN expression within these lesions, which are associated with AMD disease progression, were found to co-localize with abnormal calcium deposition. Additionally, OPN puncta colocalized with an AMD risk-associated complement pathway protein, but not with apolipoprotein E or vitronectin, two other well-established basal deposit components. Mechanistically, we found that retinal pigment epithelial cells, cells vulnerable in AMD, will secrete OPN into the extracellular space, under oxidative stress conditions, supporting OPN biosynthesis locally within the outer retina. Finally, we report that OPN levels in plasma of aged (non-AMD) human donors were significantly higher than levels in young (non-AMD) donors, but were not significantly different from donors with the different clinical subtypes of AMD. Collectively, our study defines the expression pattern of OPN in the posterior pole as a function of disease, and its local expression as a potential histopathologic biomarker of AMD.

Role of Optical Coherence Tomography Imaging in Predicting Progression of Age-Related Macular Disease: A Survey

In developed countries, age-related macular degeneration (AMD), a retinal disease, is the main cause of vision loss in the elderly. Optical Coherence Tomography (OCT) is currently the gold standard for assessing individuals for initial AMD diagnosis. In this paper, we look at how OCT imaging can be used to diagnose AMD. Our main aim is to examine and compare automated computer-aided diagnostic (CAD) systems for diagnosing and grading of AMD. We provide a brief summary, outlining the main aspects of performance assessment and providing a basis for current research in AMD diagnosis. As a result, the only viable alternative is to prevent AMD and stop both this devastating eye condition and unwanted visual impairment. On the other hand, the grading of AMD is very important in order to detect early AMD and prevent patients from reaching advanced AMD disease. In light of this, we explore the remaining issues with automated systems for AMD detection based on OCT imaging, as well as potential directions for diagnosis and monitoring systems based on OCT imaging and telemedicine applications.

Inner Retinal Layer Thickness Alterations in Early Age Related Macular Degeneration in Eyes with Subretinal Drusenoid Deposits or Conventional Drusen

The purpose of this study was to evaluate central and parafoveal inner retinal layer thickness in patients with subretinal drusenoid deposits (SDD) or conventional drusen (CD). Participants underwent comprehensive ophthalmoscopic examination. Evidence of SDD or CD was evaluated with near infrared reflectance and spectral domain optical coherence tomography. Quantification of subfoveal lesions was made through a qualitative analysis of vertical and horizontal SD-OCT scans centered on the fovea. Inner retinal layer macular thickness measurements were obtained for central circles with 1, 3, and 5 mm diameter. Continuous variables were compared by the analysis of covariance (ANCOVA) with post-hoc Tukey HSD correction for multiple comparison analysis. Fifty-five patients were included in the study; 18 eyes with SDD alone, 19 eyes with CD alone, and 18 eyes of healthy age-matched subjects. Eight eyes with SDD (44%) and 13 eyes with CD (68%) had subfoveal lesions. There was significant reduction in the inner retinal layer thickness in the central 1mm area and in the superior 3 mm area in the SDD and CD group compared to controls. In conclusion the inner retinal layer is thinner in the central macula and in the superior parafovea in eyes.

Feature definition and comprehensive analysis on the robust identification of intraretinal cystoid regions using optical coherence tomography images

Currently, optical coherence tomography is one of the most used medical imaging modalities, offering cross-sectional representations of the studied tissues. This image modality is specially relevant for the analysis of the retina, since it is the internal part of the human body that allows an almost direct examination without invasive techniques. One of the most representative cases of use of this medical imaging modality is for the identification and characterization of intraretinal fluid accumulations, critical for the diagnosis of one of the main causes of blindness in developed countries: the Diabetic Macular Edema. The study of these fluid accumulations is particularly interesting, both from the point of view of pattern recognition and from the different branches of health sciences. As these fluid accumulations are intermingled with retinal tissues, they present numerous variants according to their severity, and change their appearance depending on the configuration of the device; they are a perfect subject for an in-depth research, as they are considered to be a problem without a strict solution. In this work, we propose a comprehensive and detailed analysis of the patterns that characterize them. We employed a pool of 11 different texture and intensity feature families (giving a total of 510 markers) which we have analyzed using three different feature selection strategies and seven complementary classification algorithms. By doing so, we have been able to narrow down and explain the factors affecting this kind of accumulations and tissue lesions by means of machine learning techniques with a pipeline specially designed for this purpose.

The Effect of Selective Retina Therapy for Bevacizumab-Resistant Chronic Central Serous Chorioretinopathy

PURPOSE

The aim of this study was to evaluate the efficacy of selective retina therapy (SRT), used in conjunction with real-time feedback dosimetry (RFD), in the treatment of bevacizumab-resistant chronic central serous chorioretinopathy (CSC).

PATIENTS AND METHODS

In this retrospective cohort study, 22 eyes of 22 patients with bevacizumab-resistant chronic CSC, showing focal or diffuse foveal leakages on fundus fluorescein angiography (FFA), were included. After evaluation of the test spots at temporal arcades, SRT (wavelength, 527 nm; pulse repetition rate, 100 Hz; ramping over maximal 15 micropulses; and spot diameter, 200 μm) using RFD was applied to the leakage sites observed on FFA. Changes in the mean best-corrected visual acuity (BCVA), central macular thickness (CMT), and subretinal fluid (SRF) height were evaluated at baseline and at 1, 3, 6, 9, and 12 months following treatment.

RESULTS

SRF completely resolved in 81.8% (18/22 eyes) cases at 12 months post-treatment. The mean BCVA (logarithm of the minimum angle of resolution [logMAR]) improved from 0.49 ± 0.29 at baseline to 0.43 ± 0.36 at 12 months (p = 0.067). The mean BCVA gain was 0.06 logMAR, equivalent to 3 ETDRS letters. The CMT significantly decreased from 323 ± 85.6 μm at baseline to 221.5 ± 60.4 μm at 12 months (p < 0.001). The mean SRF height also significantly decreased from 174.6 ± 86.4 μm at baseline to 35.1 ± 75.4 μm at 12 months (p < 0.001).

CONCLUSION

SRT showed favorable visual and anatomical outcomes in patients with bevacizumab-resistant chronic CSC.

Ultra-widefield and anterior-segment optical coherence tomography in Alagille syndrome

Purpose

We describe the ocular findings in a patient with Alagille syndrome, included those obtained with ultra-widefield and anterior-segment optical coherence tomography (AS-OCT) imaging.

Observations

A previously asymptomatic 29-year-old woman with a heterozygous pathogenic variant in the JAG1 gene was referred for an ophthalmic evaluation. The ocular abnormalities included bilateral posterior embryotoxon, iris atrophy, retinal pigmentary changes in the peripheral and peripapillary regions, and optic disc elevation. Ultra-widefield OCT showed bilateral retinal thinning with increased choroidal hyperreflectivity in the areas of peripheral retinopathy and optic disc elevation. AS-OCT confirmed bilateral iris atrophy.

Conclusions and Importance

The ocular abnormalities observed in the present case represent clinical features characteristic of Alagille syndrome. Both ultra-widefield and AS-OCT were useful for assessing the ocular abnormalities in Alagille syndrome.

Predominantly Persistent Subretinal Fluid in the Comparison of Age-Related Macular Degeneration Treatments Trials

OBJECTIVE

To describe predominantly persistent subretinal fluid (SRF) in eyes receiving ranibizumab or bevacizumab for neovascular age-related macular degeneration and to compare visual acuity (VA) to eyes with nonpersistent SRF.

DESIGN

Cohort within randomized clinical trial.

PARTICIPANTS

Comparison of Age-related Macular Degeneration Treatments Trials patients assigned to pro re nata treatment.

METHODS

Graders evaluated monthly OCT scans for SRF. Predominantly persistent SRF through week 12 was defined as SRF at baseline and weeks 4, 8, and 12. Predominantly persistent SRF through 1 or 2 years was defined as SRF in 80% or more of visits by years 1 or 2, respectively. Linear regression models including baseline predictors of VA and predominantly persistent intraretinal fluid (IRF) were used to evaluate mean differences in vision outcomes.

PRIMARY OUTCOME MEASURES

Predominantly persistent SRF through year 1, adjusted VA score and VA change, and foveal SRF thickness.

RESULTS

Among 406 eyes with baseline SRF, SRF persisted in 108 eyes (26.6%) through week 12, in 94 eyes (23.2%) through year 1, and in 77 eyes (19.0%) through year 2. Adjusted VA means at year 1 were similar between eyes with predominantly persistent versus non persistent SRF by week 12 (68.1 vs. 70.2 letters; P = 0.18), year 1 (67.6 vs. 70.2 letters; P = 0.11), and year 2 (71.4 vs. 70.9 letters; P = 0.78). Adjusted changes in mean VA at year 1 were similar between eyes with predominantly persistent versus nonpersistent SRF by week 12 (6.3 vs. 7.6 letters; P = 0.38), year 1 (5.5 vs. 7.8 letters; P = 0.14), and year 2 (8.1 vs. 7.7 letters; P = 0.78). Among eyes with predominantly persistent SRF through year 1, foveal SRF was absent in 46 eyes (48.9%), thickness was 1 to 200 μm in 48 eyes (50.0%) and more than 200 μm in 1 eye (1.1%) at year 1.

CONCLUSIONS

Eyes with predominantly persistent and nonpersistent SRF through week 12, year 1, or year 2 showed similar VA outcomes after adjustment for baseline covariates and persistent IRF. At the foveal center, predominantly persistent SRF was most commonly absent or present in small quantities.

Prechoroidal cleft thickness correlates with disease activity in neovascular age-related macular degeneration

Purpose

The purpose of this study was to investigate the structural variations of the hyporeflective pocket of fluid (prechoroidal cleft) located between Bruch’s membrane and the hyperreflective material within the pigment epithelial detachment (PED) in patients with neovascular age-related macular degeneration (nAMD).

Methods

In this retrospective, observational case series study, patients diagnosed with nAMD and prechoroidal cleft associated with other activity signs of the macular neovascularization (MNV) were included. Structural optical coherence tomography (OCT) scans were evaluated to obtain anatomical measurements of prechoroidal cleft and PED at three different visits (T0, inactive MNV; T1, active MNV; T2, treated inactive MNV). The variations in size of the cleft and the PED were correlated with nAMD activity.

Results

Twenty-nine eyes from 27 patients were included. The subfoveal measurements showed a significant increase of prechoroidal cleft height and width from T0 to T1 (P < 0.05) and a subsequent decrease of the cleft height after treatment with anti-VEGF agents (P = 0.004). A similar significant trend was observed for the greatest prechoroidal cleft height and width, obtained assessing the whole OCT raster. In the multivariate analysis, the cleft height was significantly affected by both time (P = 0.001) and PED height (P < 0.0001). By contrast, the effect of fibrovascular tissue size within the PED was not significant. Visual acuity did not correlate with prechoroidal cleft size.

Conclusion

Prechoroidal cleft increased in association with MNV reactivation and decreased after treatment. Our results suggest that prechoroidal cleft could represent an accumulation of fluid actively exudating from the MNV and should be considered a sign of nAMD activity.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00417-021-05384-w.

Hyperreflective Foci, Optical Coherence Tomography Progression Indicators in Age-Related Macular Degeneration, Include Transdifferentiated Retinal Pigment Epithelium

Purpose

By optical coherence tomography (OCT) imaging, hyperreflective foci (HRF) indicate progression risk for advanced age-related macular degeneration (AMD) and are in part attributable to ectopic retinal pigment epithelium (RPE). We hypothesized that ectopic RPE are molecularly distinct from in-layer cells and that their cross-retinal course follows Müller glia.

Methods

In clinical OCT (61 eyes, 44 patients with AMD, 79.4 ± 7.7 years; 29 female; follow-up = 4.7 ± 0.9 years), one HRF type, RPE plume (n = 129 in 4 morphologies), was reviewed. Twenty eyes of 20 donors characterized by ex vivo OCT were analyzed by histology (normal, 4; early/intermediate AMD, 7; geographic atrophy, 6; neovascular AMD, 3). Cryosections were stained with antibodies to retinoid (RPE65, CRALPB) and immune (CD68, CD163) markers. In published RPE cellular phenotypes, red immunoreactivity was assessed semiquantitatively by one observer (none, some cells, all cells).

Results

Plume morphology evolved over time and many resolved (40%). Trajectories of RPE plume and cellular debris paralleled Müller glia, including near atrophy borders. RPE corresponding to HRF lost immunoreactivity for retinoid markers and gained immunoreactivity for immune markers. Aberrant immunoreactivity appeared in individual in-layer RPE cells and extended to all abnormal phenotypes. Müller glia remained CRALBP positive. Plume cells approached and contacted retinal capillaries.

Conclusions

HRF are indicators not predictors of overall disease activity. Gain and loss of function starts with individual in-layer RPE cells and extends to all abnormal phenotypes. Evidence for RPE transdifferentiation, possibly due to ischemia, supports a proposed process of epithelial–mesenchyme transition. Data can propel new biomarkers and therapeutic strategies for AMD.

Validation and Clinical Applicability of Whole-Volume Automated Segmentation of Optical Coherence Tomography in Retinal Disease Using Deep Learning

Question

Is deep learning–based segmentation of macular disease in optical coherence tomography (OCT) suitable for clinical use?

Findings

In this diagnostic study of OCT data from 173 patients with age-related macular degeneration or diabetic macular edema, model segmentations qualitatively ranked better or comparable for clinical applicability to 1 or more expert grader segmentations in 127 scans (73%) by a panel of 3 retinal specialists. Scans with high quantitative accuracy scores were not reliably associated with higher rankings.

Meaning

These findings suggest that qualitative evaluation adds to quantitative approaches when assessing clinical applicability of segmentation tools and clinician satisfaction in practice.

Importance

Quantitative volumetric measures of retinal disease in optical coherence tomography (OCT) scans are infeasible to perform owing to the time required for manual grading. Expert-level deep learning systems for automatic OCT segmentation have recently been developed. However, the potential clinical applicability of these systems is largely unknown.

Objective

To evaluate a deep learning model for whole-volume segmentation of 4 clinically important pathological features and assess clinical applicability.

Design, Setting, Participants

This diagnostic study used OCT data from 173 patients with a total of 15 558 B-scans, treated at Moorfields Eye Hospital. The data set included 2 common OCT devices and 2 macular conditions: wet age-related macular degeneration (107 scans) and diabetic macular edema (66 scans), covering the full range of severity, and from 3 points during treatment. Two expert graders performed pixel-level segmentations of intraretinal fluid, subretinal fluid, subretinal hyperreflective material, and pigment epithelial detachment, including all B-scans in each OCT volume, taking as long as 50 hours per scan. Quantitative evaluation of whole-volume model segmentations was performed. Qualitative evaluation of clinical applicability by 3 retinal experts was also conducted. Data were collected from June 1, 2012, to January 31, 2017, for set 1 and from January 1 to December 31, 2017, for set 2; graded between November 2018 and January 2020; and analyzed from February 2020 to November 2020.

Main Outcomes and Measures

Rating and stack ranking for clinical applicability by retinal specialists, model-grader agreement for voxelwise segmentations, and total volume evaluated using Dice similarity coefficients, Bland-Altman plots, and intraclass correlation coefficients.

Results

Among the 173 patients included in the analysis (92 [53%] women), qualitative assessment found that automated whole-volume segmentation ranked better than or comparable to at least 1 expert grader in 127 scans (73%; 95% CI, 66%-79%). A neutral or positive rating was given to 135 model segmentations (78%; 95% CI, 71%-84%) and 309 expert gradings (2 per scan) (89%; 95% CI, 86%-92%). The model was rated neutrally or positively in 86% to 92% of diabetic macular edema scans and 53% to 87% of age-related macular degeneration scans. Intraclass correlations ranged from 0.33 (95% CI, 0.08-0.96) to 0.96 (95% CI, 0.90-0.99). Dice similarity coefficients ranged from 0.43 (95% CI, 0.29-0.66) to 0.78 (95% CI, 0.57-0.85).

Conclusions and Relevance

This deep learning–based segmentation tool provided clinically useful measures of retinal disease that would otherwise be infeasible to obtain. Qualitative evaluation was additionally important to reveal clinical applicability for both care management and research.

Identification and quantification of fibrotic areas in the human retina using polarization-sensitive OCT

Subretinal fibrosis is one of the most prevalent causes of blindness in the elderly population, but a true gold standard to objectively diagnose fibrosis is still lacking. Since fibrotic tissue is birefringent, it can be detected by polarization-sensitive optical coherence tomography (PS-OCT). We present a new algorithm to automatically detect, segment, and quantify fibrotic lesions within 3D data sets recorded by PS-OCT. The algorithm first compensates for the birefringence of anterior ocular tissues and then uses the uniformity of the birefringent optic axis as an indicator to identify fibrotic tissue, which is then segmented and quantified. The algorithm was applied to 3D volumes recorded in 57 eyes of 57 patients with neovascular age-related macular degeneration using a spectral domain PS-OCT system. The results of fibrosis detection were compared to the clinical diagnosis based on color fundus photography (CFP), and the precision of fibrotic area measurement was assessed by three repeated measurements in a sub-set of 15 eyes. The average standard deviation of the fibrotic area obtained in eyes with a lesion area > 0.7 mm2 was 15%. Fibrosis detection by CFP and PS-OCT agreed in 48 cases, discrepancies were only observed in cases of lesion area < 0.7 mm2. These remaining discrepancies are discussed, and a new method to treat ambiguous cases is presented.

USE OF CHOROIDAL VASCULARITY INDEX FOR CHOROIDAL STRUCTURAL EVALUATION IN CENTRAL SEROUS CHORIORETINOPATHY WITH CHOROIDAL NEOVASCULARIZATION

PURPOSE

To evaluate choroidal vascular structure in eyes with central serous chorioretinopathy (CSC) by assessing the choroidal vascular index (CVI).

METHODS

We retrospectively analyzed the medical records of 117 eyes with CSC. Subjects were divided into 4 groups according to clinical characteristics: 1) acute CSC (N = 29), 2) non-neovascularized chronic CSC without flat irregular pigment epithelial detachment (N = 49), 3) non-neovascularized chronic CSC with flat irregular pigment epithelial detachment (N = 21), and 4) chronic CSC with choroidal neovascularization (N = 18). Subfoveal choroidal area (1,500 mm) of swept source optical coherence tomography scans were divided into luminal and stromal areas by the image binarization technique. The CVI was defined as the ratio of the luminal to the total subfoveal choroidal area.

RESULTS

The CVI was significantly lower in eyes of Group 4 than those of other groups (all P < 0.05). The subfoveal choroidal thickness was significantly lower in Group 4 than in Groups 1 and 2 (P < 0.05), but regression analysis showed no association with the CVI.

CONCLUSION

Decreased CVI may reflect choroidal vascular structure changes in eyes with choroidal neovascularization complicating CSC. These findings suggest that the CVI could be useful for evaluating choroidal vascular changes in eyes with CSC.

Association of Macular Thickness With Age and Age-Related Macular Degeneration in the Carotenoids in Age-Related Eye Disease Study 2 (CAREDS2), An Ancillary Study of the Women's Health Initiative

Purpose

To evaluate the relationship of retinal layer thickness with age and age-related macular degeneration (AMD) in the Carotenoids in Age-Related Eye Disease Study 2.

Methods

Total retinal thickness within the macular area, and individual layer thickness was determined for CAREDS2 participants (n = 906 eyes, 473 women) from the Women's Health Initiative using Heidelberg optical coherence tomography (OCT). Mean measurements within the OCT grid were compared across age tertiles (69–78, 78–83, and 83–101 years) and AMD outcomes.

Results

Mean retinal thickness in the central circle, inner ring, and outer ring were 277 ± 34 µm, 326 ± 20 µm, and 282 ± 15 µm, respectively. Thickness did not vary by age in the central circle, but decreased with age in the inner and outer circles (P ≤ 0.004). Specifically, ganglion cell (GCL), inner plexiform, and outer nuclear (ONL) layer thickness decreased with age (P ≤ 0.003). Age-adjusted retinal thickness in all three circles did not vary by AMD outcomes (486 without AMD and 413 with AMD). However, individual layers showed changes with GCL and photoreceptor thinning and retinal pigment epithelial thicknening in eyes with late AMD. After controlling for age and AMD, higher ONL thickness was associated with better visual acuity.

Conclusions

In this cohort of older women, a decrease in perifoveal thickness was associated with increasing age, particularly in the inner retinal layers. Variabilty in thickness in AMD eyes was primarily due to outer retinal layers. Among all retinal layers, the ONL plays an important role in preserving visual acuity.

Translational Relevance

The study provides a deeper understanding of age related changes to the retinal layers and their effect on visual loss.

Assessing the Clinical Utility of Expanded Macular OCTs Using Machine Learning

Purpose

Optical coherence tomography (OCT) is widely used in the management of retinal pathologies, including age-related macular degeneration (AMD), diabetic macular edema (DME), and primary open-angle glaucoma (POAG). We used machine learning techniques to understand diagnostic performance gains from expanding macular OCT B-scans compared with foveal-only OCT B-scans for these conditions.

Methods

Electronic medical records were extracted to obtain 61 B-scans per eye from patients with AMD, diabetic retinopathy, or POAG. We constructed deep neural networks and random forest ensembles and generated area under the receiver operating characteristic (AUROC) and area under the precision recall (AUPR) curves.

Results

After extracting 630,000 OCT images, we achieved improved AUROC and AUPR curves when comparing the central image (one B-scan) to all images (61 B-scans). The AUROC and AUPR points of diminishing return for diagnostic accuracy for macular OCT coverage were found to be within 2.75 to 4.00 mm (14–19 B-scans), 4.25 to 4.50 mm (20–21 B-scans), and 4.50 to 6.25 mm (21–28 B-scans) for AMD, DME, and POAG, respectively. All models with >0.25 mm of coverage had statistically significantly improved AUROC/AUPR curves for all diseases (P < 0.05).

Conclusions

Systematically expanded macular coverage models demonstrated significant differences in total macular coverage required for improved diagnostic accuracy, with the largest macular area being relevant in POAG followed by DME and then AMD. These findings support our hypothesis that the extent of macular coverage by OCT imaging in the clinical setting, for any of the three major disorders, has a measurable impact on the functionality of artificial intelligence decision support.

Translational Relevance

We used machine learning techniques to improve OCT imaging standards for common retinal disease diagnoses.

Quantification of Key Retinal Features in Early and Late Age-Related Macular Degeneration Using Deep Learning

PURPOSE

We sought to develop and validate a deep learning model for segmentation of 13 features associated with neovascular and atrophic age-related macular degeneration (AMD).

DESIGN

Development and validation of a deep-learning model for feature segmentation.

METHODS

Data for model development were obtained from 307 optical coherence tomography volumes. Eight experienced graders manually delineated all abnormalities in 2712 B-scans. A deep neural network was trained with these data to perform voxel-level segmentation of the 13 most common abnormalities (features). For evaluation, 112 B-scans from 112 patients with a diagnosis of neovascular AMD were annotated by 4 independent observers. The main outcome measures were Dice score, intraclass correlation coefficient, and free-response receiver operating characteristic curve.

RESULTS

On 11 of 13 features, the model obtained a mean Dice score of 0.63 ± 0.15, compared with 0.61 ± 0.17 for the observers. The mean intraclass correlation coefficient for the model was 0.66 ± 0.22, compared with 0.62 ± 0.21 for the observers. Two features were not evaluated quantitatively because of a lack of data. Free-response receiver operating characteristic analysis demonstrated that the model scored similar or higher sensitivity per false positives compared with the observers.

CONCLUSIONS

The quality of the automatic segmentation matches that of experienced graders for most features, exceeding human performance for some features. The quantified parameters provided by the model can be used in the current clinical routine and open possibilities for further research into treatment response outside clinical trials.

Brolucizumab vs aflibercept and ranibizumab for neovascular age-related macular degeneration: a cost-effectiveness analysis

BACKGROUND: Age-related macular degeneration (AMD) is a leading cause of blindness worldwide and is the most common cause of blindness in developed countries. Despite antivascular endothelial growth factor (anti-VEGF) therapy demonstrating improvements in visual and anatomical outcomes, unmet needs remain. Brolucizumab-dbll (ie, brolucizumab), a VEGF inhibitor for treatment of neovascular (wet) AMD and recently approved by the FDA for its treatment of wet AMD, attempts to mitigate treatment burden through less frequent injections. OBJECTIVE: To assess the incremental cost-effectiveness of brolucizumab compared with aflibercept and ranibizumab, given similar costs per injection and the potential for longer dosing intervals based on phase 3 clinical trial data. METHODS: A Markov model was developed to model the treatment of wet AMD patients with brolucizumab vs aflibercept and vs ranibizumab over a lifetime time horizon (base case) and 5-year time horizon (scenario analysis). The Markov model consisted of 3 primary health states: on treatment, off treatment, and death. Markov substates (5 total) described visual acuity (VA) ranging from no vision impairment to blindness. These VA-based substates were defined by best-corrected visual acuity (BCVA) values measured using Early Treatment Diabetic Retinopathy Study letters. Fixed-dosing regimens for each therapy were included in the model: dosing every 4 weeks (q4w) for the first 3 months followed by dosing q8w/q12w for brolucizumab, dosing q4w for the first 3 months followed by dosing q8w for aflibercept, and q4w for ranibizumab. RESULTS: In the base case, brolucizumab was less costly than aflibercept ($63,614 vs $72,189), and brolucizumab generated 0.0079 more quality-adjusted life-years (QALYs) than aflibercept (4.580 vs 4.572). Lower total costs with brolucizumab were driven by reduced drug costs ($56,432 vs $64,057), reduced administration costs ($6,013 vs $6,825), and reduced monitoring costs ($1,168 vs $1,306). When evaluating the cost-effectiveness of brolucizumab over a 5-year time horizon, brolucizumab was less costly than aflibercept ($44,644 vs $50,772) and generated an additional 0.0049 QALYs (2.953 vs 2.948). Additionally, brolucizumab was less costly than ranibizumab ($63,614 vs $128,163) and generated 0.0078 more QALYs than ranibizumab (4.580 vs 4.572) in the base case. Lower total costs with brolucizumab were driven by reduced drug costs ($56,432 vs $114,516), reduced administration costs ($6,013 vs $11,541), and reduced monitoring costs ($1,168 vs $2,107). When evaluating the cost-effectiveness of brolucizumab over a 5-year time horizon, brolucizumab was less costly than ranibizumab ($44,644 vs $89,665), and brolucizumab generated an additional 0.0046 QALYs (2.953 vs 2.948). CONCLUSIONS: Brolucizumab can be cost saving and cost-effective compared with aflibercept and ranibizumab in the treatment of wet AMD. DISCLOSURES: Novartis Pharmaceuticals Corporation provided funding to Xcenda for the cost-effectiveness analysis and preparation of this manuscript. Carlton is an employee of Xcenda. Agashivala is employed by Novartis Pharmaceuticals Corporation; Yu was an employee of Novartis Pharmaceutical Corporation at the time of this study. Hassan reports personal fees from iOPEN, BVI/Visitrec, ArcticDx, Bayer, F. Hoffmann-La Roche Ltd, Broadspot, BMC, Katalyst Surgical, Alcon, Vitreq, Surgicube, personal Ocugenix, Regeneron, Allergan, Oculus Surgical, Novartis, Genentech, and Eyepoint, unrelated to this work. Wykoff reports personal fees from Corcept Therapeutics, DORC, EyePoint, Gyroscope, IVERIC Bio, Merck, Notal Vision, ONL Therapeutics, Oxurion, Palatin, PolyPhotonix, Takeda, Thea Open Innovation; grants from Aerie Pharmaceuticals, Aldeyra, Gemini Therapeutics, Graybug Vision, IONIS Pharmaceutical, LMRI, Mylan, Neurotech Pharmaceuticals, Outlook Pharmaceuticals, Samsung Bioepis, Senju, Taiwan Liposome Company, Xbrane BioPharma, Santen; and grants and personal fees from Adverum, Allergan, Apellis, Chengdu Kanghong Biotechnologies (KHB), Clearside Biomedical, Genentech, Kodiak Sciences, NGM Biopharmaceuticals, Novartis, Opthea, Recens Medical, Regenxbio, Roche, and Regeneron, unrelated to this work. This research was presented as a virtual poster at the AMCP 2020 Annual Meeting, April 2020.

Automated identification of clinical features from sparsely annotated 3-dimensional medical imaging

One of the core challenges in applying machine learning and artificial intelligence to medicine is the limited availability of annotated medical data. Unlike in other applications of machine learning, where an abundance of labeled data is available, the labeling and annotation of medical data and images require a major effort of manual work by expert clinicians who do not have the time to annotate manually. In this work, we propose a new deep learning technique (SLIVER-net), to predict clinical features from 3-dimensional volumes using a limited number of manually annotated examples. SLIVER-net is based on transfer learning, where we borrow information about the structure and parameters of the network from publicly available large datasets. Since public volume data are scarce, we use 2D images and account for the 3-dimensional structure using a novel deep learning method which tiles the volume scans, and then adds layers that leverage the 3D structure. In order to illustrate its utility, we apply SLIVER-net to predict risk factors for progression of age-related macular degeneration (AMD), a leading cause of blindness, from optical coherence tomography (OCT) volumes acquired from multiple sites. SLIVER-net successfully predicts these factors despite being trained with a relatively small number of annotated volumes (hundreds) and only dozens of positive training examples. Our empirical evaluation demonstrates that SLIVER-net significantly outperforms standard state-of-the-art deep learning techniques used for medical volumes, and its performance is generalizable as it was validated on an external testing set. In a direct comparison with a clinician panel, we find that SLIVER-net also outperforms junior specialists, and identifies AMD progression risk factors similarly to expert retina specialists.

Application of Deep Learning for Diagnosing, Classifying, and Treating Age-Related Macular Degeneration

Age-related macular degeneration (AMD) affects nearly 200 million people and is the third leading cause of irreversible vision loss worldwide. Deep learning, a branch of artificial intelligence that can learn image recognition based on pre-existing datasets, creates an opportunity for more accurate and efficient diagnosis, classification, and treatment of AMD on both individual and population levels. Current algorithms based on fundus photography and optical coherence tomography imaging have already achieved diagnostic accuracy levels comparable to human graders. This accuracy can be further increased when deep learning algorithms are simultaneously applied to multiple diagnostic imaging modalities. Combined with advances in telemedicine and imaging technology, deep learning can enable large populations of patients to be screened than would otherwise be possible and allow ophthalmologists to focus on seeing those patients who are in need of treatment, thus reducing the number of patients with significant visual impairment from AMD.

Visual outcomes following cataract surgery in age-related macular degeneration patients

OBJECTIVE

To evaluate visual acuity (VA) outcomes, prognostic factors, and changes in disease severity in patients with age-related macular degeneration (AMD) undergoing cataract surgery.

DESIGN

Retrospective cohort study PARTICIPANTS: Patients with AMD or healthy control patients who underwent cataract surgery between 2012 and 2017.

METHODS

Eyes were categorized into 3 AMD groups-intermediate AMD (iAMD), fovea-involving geographic atrophy (GA), neovascular AMD (nAMD)-and 3 preoperative VA-matched control groups (iAMDc), fovea-involving geographic atrophy control (GAc), neovaascular AMD control (nAMDc).

RESULTS

We compared 216 iAMD, 35 GA, and 184 nAMD eyes with 130, 31, and 129 controls. At postoperative month 12 (POM12), VA increased significantly in iAMD and nAMD (+10.1 ± 14.5 and +9.7 ± 18.9 letters, p < 0.001), but not in GA (p = 0.68). All control groups showed significant VA gains (iAMDc: +17.1 ± 9.7, GAc: +30 ± 12.9, and nAMDc: +26.4 ± 15.6 letters, p < 0.001). For AMD groups, POM12 VA and gain in VA were significantly lower than that of controls (p < 0.01), and better preoperative VA predicted smaller VA gains (p ≤ 0.007). Longer duration of AMD in iAMD, ellipsoid zone disruption in nAMD, and lower central subfield thickness in GA were associated with poorer VA outcomes (p < 0.05). Development of nAMD occurred in 8 iAMD eyes and was associated with longer duration of disease (p = 0.001). For nAMD eyes, injection frequency did not vary between the 12-month pre- and postoperative periods (p = 0.051).

CONCLUSIONS

Cataract surgery improves VA for patients with iAMD and nAMD, albeit not to the level of those without retinal pathology. Preoperative VA, AMD duration, and optical coherence tomography parameters may be important prognostic factors for cataract surgery in patients with AMD.