Deep Learning-Based Prediction of Individual Geographic Atrophy Progression from a Single Baseline OCT

Objective

To identify the individual progression of geographic atrophy (GA) lesions from baseline OCT images of patients in routine clinical care.

Design

Clinical evaluation of a deep learning-based algorithm.

Subjects

One hundred eighty-four eyes of 100 consecutively enrolled patients.

Methods

OCT and fundus autofluorescence (FAF) images (both Spectralis, Heidelberg Engineering) of patients with GA secondary to age-related macular degeneration in routine clinical care were used for model validation. Fundus autofluorescence images were annotated manually by delineating the GA area by certified readers of the Vienna Reading Center. The annotated FAF images were anatomically registered in an automated manner to the corresponding OCT scans, resulting in 2-dimensional en face OCT annotations, which were taken as a reference for the model performance. A deep learning-based method for modeling the GA lesion growth over time from a single baseline OCT was evaluated. In addition, the ability of the algorithm to identify fast progressors for the top 10%, 15%, and 20% of GA growth rates was analyzed.

Main Outcome Measures

Dice similarity coefficient (DSC) and mean absolute error (MAE) between manual and predicted GA growth.

Results

The deep learning-based tool was able to reliably identify disease activity in GA using a standard OCT image taken at a single baseline time point. The mean DSC for the total GA region increased for the first 2 years of prediction (0.80-0.82). With increasing time intervals beyond 3 years, the DSC decreased slightly to a mean of 0.70. The MAE was low over the first year and with advancing time slowly increased, with mean values ranging from 0.25 mm to 0.69 mm for the total GA region prediction. The model achieved an area under the curve of 0.81, 0.79, and 0.77 for the identification of the top 10%, 15%, and 20% growth rates, respectively.

Conclusions

The proposed algorithm is capable of fully automated GA lesion growth prediction from a single baseline OCT in a time-continuous fashion in the form of en face maps. The results are a promising step toward clinical decision support tools for therapeutic dosing and guidance of patient management because the first treatment for GA has recently become available.

Financial Disclosures

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

A Systematic Prospective Comparison of Fluid Volume Evaluation across OCT Devices Used in Clinical Practice

Objective

Treatment decisions in neovascular age-related macular degeneration (nAMD) are mainly based on subjective evaluation of OCT. The purpose of this cross-sectional study was to provide a comparison of qualitative and quantitative differences between OCT devices in a systematic manner.

Design

Prospective, cross-sectional study.

Subjects

One hundred sixty OCT volumes, 40 eyes of 40 patients with nAMD.

Methods

Patients from clinical practice were imaged with 4 different OCT devices during one visit: (1) Spectralis Heidelberg; (2) Cirrus; (3) Topcon Maestro2; and (4) Topcon Triton. Intraretinal fluid (IRF), subretinal fluid (SRF), and pigment epithelial detachment (PED) were manually annotated in all cubes by trained human experts to establish fluid measurements based on expert-reader annotations. Intraretinal fluid, SRF, and PED volume were quantified in nanoliters (nL). Bland-Altman plots were created to analyze the agreement of measurements in the central 1 and 6 mm. The Friedman test was performed to test for significant differences in the central 1, 3, and 6 mm.

Main Outcome Measures

Intraretinal fluid, SRF, and PED volume.

Results

In the central 6 mm, there was a trend toward higher IRF and PED volumes in Spectralis images compared with the other devices and no differences in SRF volume. In the central 1 mm, the standard deviation of the differences ranged from ± 3 nL to ± 6 nL for IRF, from ± 3 nL to ± 4 nL for SRF, and from ± 7 nL to ± 10 nL for PED in all pairwise comparisons. Manually annotated IRF and SRF volumes showed no significant differences in the central 1 mm.

Conclusions

Fluid volume quantification achieved excellent reliability in all 3 retinal compartments on images obtained from 4 OCT devices, particularly for clinically relevant IRF and SRF values. Although fluid volume quantification is reliable in all 4 OCT devices, switching OCT devices might lead to deviating fluid volume measurements with higher agreement in the central 1 mm compared with the central 6 mm, with highest agreement for SRF volume in the central 1 mm. Understanding device-dependent differences is essential for expanding the interpretation and implementation of pixel-wise fluid volume measurements in clinical practice and in clinical trials.

Financial Disclosures

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

3DTINC: Time-Equivariant Non-Contrastive Learning for Predicting Disease Progression from Longitudinal OCTs

Self-supervised learning (SSL) has emerged as a powerful technique for improving the efficiency and effectiveness of deep learning models. Contrastive methods are a prominent family of SSL that extract similar representations of two augmented views of an image while pushing away others in the representation space as negatives. However, the state-of-the-art contrastive methods require large batch sizes and augmentations designed for natural images that are impractical for 3D medical images. To address these limitations, we propose a new longitudinal SSL method, 3DTINC, based on non-contrastive learning. It is designed to learn perturbation-invariant features for 3D optical coherence tomography (OCT) volumes, using augmentations specifically designed for OCT. We introduce a new non-contrastive similarity loss term that learns temporal information implicitly from intra-patient scans acquired at different times. Our experiments show that this temporal information is crucial for predicting progression of retinal diseases, such as age-related macular degeneration (AMD). After pretraining with 3DTINC, we evaluated the learned representations and the prognostic models on two large-scale longitudinal datasets of retinal OCTs where we predict the conversion to wet-AMD within a six-month interval. Our results demonstrate that each component of our contributions is crucial for learning meaningful representations useful in predicting disease progression from longitudinal volumetric scans.

Morph-SSL: Self-Supervision with Longitudinal Morphing for Forecasting AMD Progression from OCT Volumes

The lack of reliable biomarkers makes predicting the conversion from intermediate to neovascular age-related macular degeneration (iAMD, nAMD) a challenging task. We develop a Deep Learning (DL) model to predict the future risk of conversion of an eye from iAMD to nAMD from its current OCT scan. Although eye clinics generate vast amounts of longitudinal OCT scans to monitor AMD progression, only a small subset can be manually labeled for supervised DL. To address this issue, we propose Morph-SSL, a novel Self-supervised Learning (SSL) method for longitudinal data. It uses pairs of unlabelled OCT scans from different visits and involves morphing the scan from the previous visit to the next. The Decoder predicts the transformation for morphing and ensures a smooth feature manifold that can generate intermediate scans between visits through linear interpolation. Next, the Morph-SSL trained features are input to a Classifier which is trained in a supervised manner to model the cumulative probability distribution of the time to conversion with a sigmoidal function. Morph-SSL was trained on unlabelled scans of 399 eyes (3570 visits). The Classifier was evaluated with a five-fold cross-validation on 2418 scans from 343 eyes with clinical labels of the conversion date. The Morph-SSL features achieved an AUC of 0.779 in predicting the conversion to nAMD within the next 6 months, outperforming the same network when trained end-to-end from scratch or pre-trained with popular SSL methods. Automated prediction of the future risk of nAMD onset can enable timely treatment and individualized AMD management.

AI-based support for optical coherence tomography in age-related macular degeneration

Artificial intelligence (AI) has emerged as a transformative technology across various fields, and its applications in the medical domain, particularly in ophthalmology, has gained significant attention. The vast amount of high-resolution image data, such as optical coherence tomography (OCT) images, has been a driving force behind AI growth in this field. Age-related macular degeneration (AMD) is one of the leading causes for blindness in the world, affecting approximately 196 million people worldwide in 2020. Multimodal imaging has been for a long time the gold standard for diagnosing patients with AMD, however, currently treatment and follow-up in routine disease management are mainly driven by OCT imaging. AI-based algorithms have by their precision, reproducibility and speed, the potential to reliably quantify biomarkers, predict disease progression and assist treatment decisions in clinical routine as well as academic studies. This review paper aims to provide a summary of the current state of AI in AMD, focusing on its applications, challenges, and prospects.

Analysing early changes of photoreceptor layer thickness following surgery in eyes with epiretinal membranes

BACKGROUND/OBJECTIVES

To analyse short-term changes of mean photoreceptor thickness (PRT) on the ETDRS-grid after vitrectomy and membrane peeling in patients with epiretinal membrane (ERM).

SUBJECTS/METHODS

Forty-eight patients with idiopathic ERM were included in this prospective study. Study examinations comprised best-corrected visual acuity (BCVA) and optical coherence tomography (OCT) before surgery, 1 week (W1), 1 month (M1) and 3 months (M3) after surgery. Mean PRT was assessed using an automated algorithm and correlated with BCVA and central retinal thickness (CRT).

RESULTS

Regarding PRT changes of the study eye in comparison to baseline values, a significant decrease at W1 in the 1 mm, 3 mm and 6 mm area (all p-values < 0.001), at M1 (p = 0.009) and M3 (p = 0.019) in the central 1 mm area, a significant increase at M3 in the 6 mm area (p < 0.001), but no significant change at M1 in the 3 mm and 6 mm area and M3 in the 3 mm area (all p-values > 0.05) were observed. BCVA increased significantly from baseline to M3 (0.3LogMAR-0.15LogMAR, Snellen equivalent = 20/40-20/28 respectively; p < 0.001). There was no correlation between baseline PRT and BCVA at any visit after surgery, nor between PRT and BCVA at any visit (all p-values > 0.05). Decrease in PRT in the 1 mm (p < 0.001), 3 mm (p = 0.013) and 6 mm (p = 0.034) area after one week correlated with the increase in CRT (449.9 µm-462.2 µm).

CONCLUSIONS

Although the photoreceptor layer is morphologically affected by ERMs and after their surgical removal, it is not correlated to BCVA. Thus, patients with photoreceptor layer alterations due to ERM may still benefit from surgery and achieve good functional rehabilitation thereafter.

Deep Multimodal Fusion of Data With Heterogeneous Dimensionality via Projective Networks

The use of multimodal imaging has led to significant improvements in the diagnosis and treatment of many diseases. Similar to clinical practice, some works have demonstrated the benefits of multimodal fusion for automatic segmentation and classification using deep learning-based methods. However, current segmentation methods are limited to fusion of modalities with the same dimensionality (e.g., 3D + 3D, 2D + 2D), which is not always possible, and the fusion strategies implemented by classification methods are incompatible with localization tasks. In this work, we propose a novel deep learning-based framework for the fusion of multimodal data with heterogeneous dimensionality (e.g., 3D + 2D) that is compatible with localization tasks. The proposed framework extracts the features of the different modalities and projects them into the common feature subspace. The projected features are then fused and further processed to obtain the final prediction. The framework was validated on the following tasks: segmentation of geographic atrophy (GA), a late-stage manifestation of age-related macular degeneration, and segmentation of retinal blood vessels (RBV) in multimodal retinal imaging. Our results show that the proposed method outperforms the state-of-the-art monomodal methods on GA and RBV segmentation by up to 3.10% and 4.64% Dice, respectively.

Learning Spatio-Temporal Model of Disease Progression With NeuralODEs From Longitudinal Volumetric Data

Robust forecasting of the future anatomical changes inflicted by an ongoing disease is an extremely challenging task that is out of grasp even for experienced healthcare professionals. Such a capability, however, is of great importance since it can improve patient management by providing information on the speed of disease progression already at the admission stage, or it can enrich the clinical trials with fast progressors and avoid the need for control arms by the means of digital twins. In this work, we develop a deep learning method that models the evolution of age-related disease by processing a single medical scan and providing a segmentation of the target anatomy at a requested future point in time. Our method represents a time-invariant physical process and solves a large-scale problem of modeling temporal pixel-level changes utilizing NeuralODEs. In addition, we demonstrate the approaches to incorporate the prior domain-specific constraints into our method and define temporal Dice loss for learning temporal objectives. To evaluate the applicability of our approach across different age-related diseases and imaging modalities, we developed and tested the proposed method on the datasets with 967 retinal OCT volumes of 100 patients with Geographic Atrophy and 2823 brain MRI volumes of 633 patients with Alzheimer's Disease. For Geographic Atrophy, the proposed method outperformed the related baseline models in the atrophy growth prediction. For Alzheimer's Disease, the proposed method demonstrated remarkable performance in predicting the brain ventricle changes induced by the disease, achieving the state-of-the-art result on TADPOLE cross-sectional prediction challenge dataset.

An Automated Comparative Analysis of the Exudative Biomarkers in Neovascular Age-Related Macular Degeneration, The RAP Study: Report 6

PURPOSE

To investigate differences in volume and distribution of the main exudative biomarkers across all types and subtypes of macular neovascularization (MNV) using artificial intelligence (AI).

DESIGN

Cross-sectional study.

METHODS

An AI-based analysis was conducted on 34,528 OCT B-scans consisting of 281 (250 unifocal, 31 multifocal) MNV3, 55 MNV2, and 121 (30 polypoidal, 91 non-polypoidal) MNV1 treatment-naive eyes. Means (SDs), medians and heat maps of cystic intraretinal fluid (IRF), subretinal fluid (SRF), pigment epithelial detachments (PED), and hyperreflective foci (HRF) volumes, as well as retinal thickness (RT) were compared among MNV types and subtypes.

RESULTS

MNV3 had the highest mean IRF with 291 (290) nL, RT with 357 (49) µm, and HRF with 80 (70) nL, P ≤ .05. MNV1 showed the greatest mean SRF with 492 (586) nL, whereas MNV3 exhibited the lowest with 218 (382) nL, P ≤ .05. Heat maps showed IRF confined to the center, whereas SRF was scattered in all types. SRF, HRF, and PED were more distributed in the temporal macular half in MNV3. Means of IRF, HRF, and PED were higher in the multifocal than in the unifocal MNV3 with 416 (309) nL,114 (95) nL, and 810 (850) nL, P ≤ .05. Compared to the non-polypoidal subtype, the polypoidal subtype had greater means of SRF with 695 (718) nL, HRF 69 (63) nL, RT 357 (45) µm, and PED 1115 (1170) nL, P ≤ .05.

CONCLUSIONS

This novel quantitative AI analysis shows that SRF is a biomarker of choroidal origin in MNV1, whereas IRF, HRF, and RT are retinal biomarkers in MNV3. Polypoidal MNV1 and multifocal MNV3 present with higher exudation compared to other subtypes.

AIROGS: Artificial Intelligence for Robust Glaucoma Screening Challenge

The early detection of glaucoma is essential in preventing visual impairment. Artificial intelligence (AI) can be used to analyze color fundus photographs (CFPs) in a cost-effective manner, making glaucoma screening more accessible. While AI models for glaucoma screening from CFPs have shown promising results in laboratory settings, their performance decreases significantly in real-world scenarios due to the presence of out-of-distribution and low-quality images. To address this issue, we propose the Artificial Intelligence for Robust Glaucoma Screening (AIROGS) challenge. This challenge includes a large dataset of around 113,000 images from about 60,000 patients and 500 different screening centers, and encourages the development of algorithms that are robust to ungradable and unexpected input data. We evaluated solutions from 14 teams in this paper and found that the best teams performed similarly to a set of 20 expert ophthalmologists and optometrists. The highest-scoring team achieved an area under the receiver operating characteristic curve of 0.99 (95% CI: 0.98-0.99) for detecting ungradable images on-the-fly. Additionally, many of the algorithms showed robust performance when tested on three other publicly available datasets. These results demonstrate the feasibility of robust AI-enabled glaucoma screening.

Point-to-point associations of drusen and hyperreflective foci volumes with retinal sensitivity in non-exudative age-related macular degeneration

OBJECTIVES

To evaluate the quantitative impact of drusen and hyperreflective foci (HRF) volumes on mesopic retinal sensitivity in non-exudative age-related macular degeneration (AMD).

METHODS

In a standardized follow-up scheme of every three months, retinal sensitivity of patients with early or intermediate AMD was assessed by microperimetry using a custom pattern of 45 stimuli (Nidek MP-3, Gamagori, Japan). Eyes were consecutively scanned using Spectralis SD-OCT (20° × 20°, 1024 × 97 × 496). Fundus photographs obtained by the MP-3 allowed to map the stimuli locations onto the corresponding OCT scans. The volume and mean thickness of drusen and HRF within a circle of 240 µm centred at each stimulus point was determined using automated AI-based image segmentation algorithms.

RESULTS

8055 individual stimuli from 179 visits from 51 eyes of 35 consecutive patients were matched with the respective OCT images in a point-to-point manner. The patients mean age was 76.85 ± 6.6 years. Mean retinal sensitivity at baseline was 25.7 dB. 73.47% of all MP-spots covered drusen area and 2.02% of MP-spots covered HRF. A negative association between retinal sensitivity and the volume of underlying drusen (p < 0.001, Estimate -0.991 db/µm3) and HRF volume (p = 0.002, Estimate -5.230 db/µm3) was found. During observation time, no eye showed conversion to advanced AMD.

CONCLUSION

A direct correlation between drusen and lower sensitivity of the overlying photoreceptors can be observed. For HRF, a small but significant correlation was shown, which is compromised by their small size. Biomarker quantification using AI-methods allows to determine the impact of sub-clinical features in the progression of AMD.

Performance of retinal fluid monitoring in OCT imaging by automated deep learning versus human expert grading in neovascular AMD

PURPOSE

To evaluate the reliability of automated fluid detection in identifying retinal fluid activity in OCT scans of patients treated with anti-VEGF therapy for neovascular age-related macular degeneration by correlating human expert and automated measurements with central retinal subfield thickness (CSFT) and fluid volume values.

METHODS

We utilized an automated deep learning approach to quantify macular fluid in SD-OCT volumes (Cirrus, Spectralis, Topcon) from patients of HAWK and HARRIER Studies. Three-dimensional volumes for IRF and SRF were measured at baseline and under therapy in the central millimeter and compared to fluid gradings, CSFT and foveal centerpoint thickness (CPT) values measured by the Vienna Reading Center.

RESULTS

41.906 SD-OCT volume scans were included into the analysis. Concordance between human expert grading and automated algorithm performance reached AUC values of 0.93/0.85 for IRF and 0.87 for SRF in HARRIER/HAWK in the central millimeter. IRF volumes showed a moderate correlation with CSFT at baseline (HAWK: r = 0.54; HARRIER: r = 0.62) and weaker correlation under therapy (HAWK: r = 0.44; HARRIER: r = 0.34). SRF and CSFT correlations were low at baseline (HAWK: r = 0.29; HARRIER: r = 0.22) and under therapy (HAWK: r = 0.38; HARRIER: r = 0.45). The residual standard error (IRF: 75.90 µm; SRF: 95.26 µm) and marginal residual standard deviations (IRF: 46.35 µm; SRF: 44.19 µm) of fluid volume were high compared to the range of CSFT values.

CONCLUSION

Deep learning-based segmentation of retinal fluid performs reliably on OCT images. CSFT values are weak indicators for fluid activity in nAMD. Automated quantification of fluid types, highlight the potential of deep learning-based approaches to objectively monitor anti-VEGF therapy.

Long-term effect of fluid volumes during the maintenance phase in neovascular age-related macular degeneration: results from Fight Retinal Blindness!

OBJECTIVE

To investigate the effect of macular fluid volumes (subretinal fluid [SRF], intraretinal fluid [IRF], and pigment epithelium detachment [PED]) after initial treatment on functional and structural outcomes in neovascular age-related macular degeneration in a real-world cohort from Fight Retinal Blindness!

METHODS

Treatment-naive neovascular age-related macular degeneration patients from Fight Retinal Blindness! (Zürich, Switzerland) were included. Macular fluid on optical coherence tomography was automatically quantified using an approved artificial intelligence algorithm. Follow-up of macular fluid, number of anti-vascular endothelial growth factor treatments, effect of fluid volumes after initial treatment (high, top 25%; low, bottom 75%) on best-corrected visual acuity, and development of macular atrophy and fibrosis was investigated over 48 months.

RESULTS

A total of 209 eyes (mean age, 78.3 years) were included. Patients with high IRF volumes after initial treatment differed by -2.6 (p = 0.021) and -7.4 letters (p = 0.007) at months 12 and 48, respectively. Eyes with high IRF received significantly more treatments (+1.6 [p < 0.001] and +5.3 [p = 0.002] at months 12 and 48, respectively). Patients with high SRF or PED had comparable best-corrected visual acuity outcomes but received significantly more treatments for SRF (+2.4 [p < 0.001] and +11.4 [p < 0.001] at months 12 and 48, respectively) and PED (+1.2 [p = 0.001] and +7.8 [p < 0.001] at months 12 and 48, respectively).

DISCUSSION

Patients with high macular fluid after initial treatment are at risk of losing vision that may not be compensable with higher treatment frequency for IRF. Higher treatment frequency for SRF and PED may result in comparable treatment outcomes. Quantification of macular fluid in all compartments is essential to detect eyes at risk of aggressive disease.

Systematic review of prognostic factors associated with progression to late age-related macular degeneration: Pinnacle study report 2

There is a need to identify accurately prognostic factors that determine the progression of intermediate to late-stage age-related macular degeneration (AMD). Currently, clinicians cannot provide individualised prognoses of disease progression. Moreover, enriching clinical trials with rapid progressors may facilitate delivery of shorter intervention trials aimed at delaying or preventing progression to late AMD. Thus, we performed a systematic review to outline and assess the accuracy of reporting prognostic factors for the progression of intermediate to late AMD. A meta-analysis was originally planned. Synonyms of AMD and disease progression were used to search Medline and EMBASE for articles investigating AMD progression published between 1991 and 2021. Initial search results included 3229 articles. Predetermined eligibility criteria were employed to systematically screen papers by two reviewers working independently and in duplicate. Quality appraisal and data extraction were performed by a team of reviewers. Only 6 studies met the eligibility criteria. Based on these articles, exploratory prognostic factors for progression of intermediate to late AMD included phenotypic features (e.g. location and size of drusen), age, smoking status, ocular and systemic co-morbidities, race, and genotype. Overall, study heterogeneity precluded reporting by forest plots and meta-analysis. The most commonly reported prognostic factors were baseline drusen volume/size, which was associated with progression to neovascular AMD, and outer retinal thinning linked to progression to geographic atrophy. In conclusion, poor methodological quality of included studies warrants cautious interpretation of our findings. Rigorous studies are warranted to provide robust evidence in the future.

Approved AI-based fluid monitoring to identify morphological and functional treatment outcomes in neovascular age-related macular degeneration in real-world routine (FRB!)

AIM

To predict antivascular endothelial growth factor (VEGF) treatment requirements, visual acuity and morphological outcomes in neovascular age-related macular degeneration (nAMD) using fluid quantification by artificial intelligence (AI) in a real-world cohort.

METHODS

Spectral-domain optical coherence tomography data of 158 treatment-naïve patients with nAMD from the Fight Retinal Blindness! registry in Zurich were processed at baseline, and after initial treatment using intravitreal anti-VEGF to predict subsequent 1-year and 4-year outcomes. Intraretinal and subretinal fluid and pigment epithelial detachment volumes were segmented using a deep learning algorithm (Vienna Fluid Monitor, RetInSight, Vienna, Austria). A predictive machine learning model for future treatment requirements and morphological outcomes was built using the computed set of quantitative features.

RESULTS

Two hundred and two eyes from 158 patients were evaluated. 107 eyes had a lower median (≤7) and 95 eyes had an upper median (≥8) number of injections in the first year, with a mean accuracy of prediction of 0.77 (95% CI 0.71 to 0.83) area under the curve (AUC). Best-corrected visual acuity at baseline was the most relevant predictive factor determining final visual outcomes after 1 year. Over 4 years, half of the eyes had progressed to macular atrophy (MA) with the model being able to distinguish MA from non-MA eyes with a mean AUC of 0.70 (95% CI 0.61 to 0.79). Prediction for subretinal fibrosis reached an AUC of 0.74 (95% CI 0.63 to 0.81).

CONCLUSIONS

The regulatory approved AI-based fluid monitoring allows clinicians to use automated algorithms in prospectively guided patient treatment in AMD. Furthermore, retinal fluid localisation and quantification can predict long-term morphological outcomes.

[Monitoring of the progression of geographic atrophy with optical coherence tomography]

With the prospect of available therapy for geographic atrophy in the near future and consequently increasing patient numbers, appropriate management strategies for the clinical practice are needed. Optical coherence tomography (OCT) as well as automated OCT analysis using artificial intelligence algorithms provide optimal conditions for assessing disease activity as well as the treatment response for geographic atrophy through a rapid, precise and resource-efficient evaluation.

Progression Dynamics of Early versus Later Stage Atrophic Lesions in Nonneovascular Age-Related Macular Degeneration Using Quantitative OCT Biomarker Segmentation

PURPOSE

To investigate the progression of geographic atrophy secondary to nonneovascular age-related macular degeneration in early and later stage lesions using artificial intelligence-based precision tools.

DESIGN

Retrospective analysis of an observational cohort study.

SUBJECTS

Seventy-four eyes of 49 patients with ≥ 1 complete retinal pigment epithelial and outer retinal atrophy (cRORA) lesion secondary to age-related macular degeneration were included. Patients were divided between recently developed cRORA and lesions with advanced disease status.

METHODS

Patients were prospectively imaged by spectral-domain OCT volume scans. The study period encompassed 18 months with scheduled visits every 6 months. Growth rates of recent cRORA-converted lesions were compared with lesions in an advanced disease status using mixed effect models.

MAIN OUTCOME MEASURES

The progression of retinal pigment epithelial loss (RPEL) was considered the primary end point. Secondary end points consisted of external limiting membrane disruption and ellipsoid zone loss. These pathognomonic imaging biomarkers were quantified using validated deep-learning algorithms. Further, the ellipsoid zone/RPEL ratio was analyzed in both study cohorts.

RESULTS

Mean (95% confidence interval [CI]) square root progression of recently converted lesions was 79.68 (95% CI, -77.14 to 236.49), 68.22 (95% CI, -101.21 to 237.65), and 84.825 (95% CI, -124.82 to 294.47) mm/half year for RPEL, external limiting membrane loss, and ellipsoid zone loss respectively. Mean square root progression of advanced lesions was 131.74 (95% CI, -22.57 to 286.05), 129.96 (95% CI, -36.67 to 296.59), and 116.84 (95% CI, -90.56 to 324.3) mm/half year for RPEL, external limiting membrane loss, and ellipsoid zone loss, respectively. RPEL (P = 0.038) and external limiting membrane disruption (P = 0.026) progression showed significant differences between the 2 study cohorts. Further recent converters had significantly (P < 0.001) higher ellipsoid zone/RPEL ratios at all time points compared with patients in an advanced disease status (1.71 95% CI, 1.12-2.28 vs. 1.14; 95% CI, 0.56-1.71).

CONCLUSION

Early cRORA lesions have slower growth rates in comparison to atrophic lesions in advanced disease stages. Differences in growth dynamics may play a crucial role in understanding the pathophysiology of nonneovascular age-related macular degeneration and for the interpretation of clinical trials in geographic atrophy. Individual disease monitoring using artificial intelligence-based quantification paves the way toward optimized geographic atrophy management.

FINANCIAL DISCLOSURE(S)

The author(s) have no proprietary or commercial interest in any materials discussed in this article.

Clinical validation for automated geographic atrophy monitoring on OCT under complement inhibitory treatment

Geographic atrophy (GA) represents a late stage of age-related macular degeneration, which leads to irreversible vision loss. With the first successful therapeutic approach, namely complement inhibition, huge numbers of patients will have to be monitored regularly. Given these perspectives, a strong need for automated GA segmentation has evolved. The main purpose of this study was the clinical validation of an artificial intelligence (AI)-based algorithm to segment a topographic 2D GA area on a 3D optical coherence tomography (OCT) volume, and to evaluate its potential for AI-based monitoring of GA progression under complement-targeted treatment. 100 GA patients from routine clinical care at the Medical University of Vienna for internal validation and 113 patients from the FILLY phase 2 clinical trial for external validation were included. Mean Dice Similarity Coefficient (DSC) was 0.86 ± 0.12 and 0.91 ± 0.05 for total GA area on the internal and external validation, respectively. Mean DSC for the GA growth area at month 12 on the external test set was 0.46 ± 0.16. Importantly, the automated segmentation by the algorithm corresponded to the outcome of the original FILLY trial measured manually on fundus autofluorescence. The proposed AI approach can reliably segment GA area on OCT with high accuracy. The availability of such tools represents an important step towards AI-based monitoring of GA progression under treatment on OCT for clinical management as well as regulatory trials.

Correlation of vascular and fluid-related parameters in neovascular age-related macular degeneration using deep learning

PURPOSE

To identify correlations between the vascular characteristics of macular neovascularization (MNV) obtained by optical coherence tomography angiography (OCTA) and distinct retinal fluid volumes in neovascular age-related macular degeneration (nAMD).

METHODS

In this prospective interventional study, 54 patients with treatment-naïve type 1 or 2 nAMD were included and treated with intravitreal aflibercept. At baseline and month 1, each patient underwent a SD-OCT volume scan and volumetric flow scan using a swept-source OCTA. A deep learning algorithm was used to automatically detect and quantify fluid in OCT scans. Angio Tool, a National Cancer Institute algorithm, was used to skeletonize MNV properties and quantify lesion size (LS), vessel area (VA), vessel density (VD), total number of endpoints (TNE), total number of junctions (TNJ), junction density (JD), total vessel length (TVL), average vessel length (AVL) and mean-e-lacunarity (MEL). Subsequently, linear regression models were used to investigate a correlation between OCTA parameters and fluid quantifications.

RESULTS

The median amount of fluid within the central 6-mm EDTRS ring was 173.7 nl at baseline, consisting of 156.6 nl of subretinal fluid (SRF) and 2.3 nl of intraretinal fluid (IRF). Fluid decreased significantly in all compartments to 1.76 nl (SRF) and 0.64 nl (IRF). The investigated MNV parameters did not change significantly after the first treatment. There was no significant correlation between MNV parameters and relative fluid decrease after anti-VEGF treatment. Baseline fluid correlated statistically significant but weakly with TNE (p = 0.002, R²  = 0.17), SRF with TVL (p = 0.04, R²  = 0.08), VD (p = 0.046, R²  = 0.08), TNE (p = 0.001, R²  = 0.20) and LS (p = 0.033, R²  = 0.09). IRF correlated with VA (p = 0.042, R²  = 0.08).The amount of IRF at month 1 correlated significantly but weakly with VD (p = 0.036, R²  = 0.08), JD (p = 0.019, R²  = 0.10) and MEL (p = 0.005, R²  = 0.14).

CONCLUSION

Macular neovascularization parameters at baseline and month 1 played only a minor role in the exudation process in nAMD. None of the MNV parameters were correlated with the treatment response.

Comparison of Fundus Autofluorescence Versus Optical Coherence Tomography-based Evaluation of the Therapeutic Response to Pegcetacoplan in Geographic Atrophy

PURPOSE

To perform an optical coherence tomography (OCT)-based analysis of geographic atrophy (GA) progression in patients treated with pegcetacoplan.

DESIGN

Post hoc analysis of a phase 2 multicenter, randomized, sham-controlled trial.

METHODS

Manual annotation of retinal pigment epithelium (RPE), ellipsoid zone (EZ), and external limiting membrane (ELM) loss was performed on OCT volumes from baseline and month 12 from the phase 2 FILLY trial of intravitreal pegcetacoplan for the treatment of GA secondary to age-related macular degeneration.

MAIN OUTCOME MEASURES

Correlation of GA areas measured on fundus autofluorescence and OCT. Difference in square root transformed growth rates of RPE, EZ, and ELM loss between treatment groups (monthly injection [AM], injection every other month [AEOM], and sham [SM]).

RESULTS

OCT volumes from 113 eyes of 113 patients (38 AM, 36 AEOM, and 39 SM) were included, resulting in 11 074 B-scans. The median growth of RPE loss was significantly slower in the AM group (0.158 [0.057-0.296]) than the SM group (0.255 [0.188-0.359], P = .014). Importantly, the growth of EZ loss was also significantly slower in the AM group (0.127 [0.041-0.247]) than the SM group (0.232 [0.130-0.349], P = .017). There was no significant difference in the growth of ELM loss between the treatment groups (P = .114).

CONCLUSIONS

OCT imaging provided consistent results for GA growth compared with fundus autofluorescence. In addition to slower RPE atrophy progression in patients treated with pegcetacoplan, a significant reduction in EZ impairment was also identified by OCT, suggesting the use of OCT as a potentially more sensitive monitoring tool in GA therapy.

ADAM Challenge: Detecting Age-Related Macular Degeneration From Fundus Images

Age-related macular degeneration (AMD) is the leading cause of visual impairment among elderly in the world. Early detection of AMD is of great importance, as the vision loss caused by this disease is irreversible and permanent. Color fundus photography is the most cost-effective imaging modality to screen for retinal disorders. Cutting edge deep learning based algorithms have been recently developed for automatically detecting AMD from fundus images. However, there are still lack of a comprehensive annotated dataset and standard evaluation benchmarks. To deal with this issue, we set up the Automatic Detection challenge on Age-related Macular degeneration (ADAM), which was held as a satellite event of the ISBI 2020 conference. The ADAM challenge consisted of four tasks which cover the main aspects of detecting and characterizing AMD from fundus images, including detection of AMD, detection and segmentation of optic disc, localization of fovea, and detection and segmentation of lesions. As part of the ADAM challenge, we have released a comprehensive dataset of 1200 fundus images with AMD diagnostic labels, pixel-wise segmentation masks for both optic disc and AMD-related lesions (drusen, exudates, hemorrhages and scars, among others), as well as the coordinates corresponding to the location of the macular fovea. A uniform evaluation framework has been built to make a fair comparison of different models using this dataset. During the ADAM challenge, 610 results were submitted for online evaluation, with 11 teams finally participating in the onsite challenge. This paper introduces the challenge, the dataset and the evaluation methods, as well as summarizes the participating methods and analyzes their results for each task. In particular, we observed that the ensembling strategy and the incorporation of clinical domain knowledge were the key to improve the performance of the deep learning models.

Fundus autofluorescence and optical coherence tomography biomarkers associated with the progression of geographic atrophy secondary to age-related macular degeneration

OBJECTIVES

To investigate the impact of qualitatively graded and deep learning quantified imaging biomarkers on growth of geographic atrophy (GA) secondary to age-related macular degeneration.

METHODS

This prospective study included 1062 visits of 181 eyes of 100 patients with GA. Spectral-domain optical coherence tomography (SD-OCT) and fundus autofluorescence (FAF) images were acquired at each visit. Hyperreflective foci (HRF) were quantitatively assessed in SD-OCT volumes using a validated deep learning algorithm. FAF images were graded for FAF patterns, subretinal drusenoid deposits (SDD), GA lesion configuration and atrophy enlargement. Linear mixed models were calculated to investigate associations between all parameters and GA progression.

RESULTS

FAF patterns were significantly associated with GA progression (p < 0.001). SDD was associated with faster GA growth (p = 0.005). Eyes with higher HRF concentrations showed a trend towards faster GA progression (p = 0.072) and revealed a significant impact on GA enlargement in interaction with FAF patterns (p = 0.01). The fellow eye status had no significant effect on lesion enlargement (p > 0.05). The diffuse-trickling FAF pattern exhibited significantly higher HRF concentrations than any other pattern (p < 0.001).

CONCLUSION

Among a wide range of investigated biomarkers, SDD and FAF patterns, particularly in interaction with HRF, significantly impact GA progression. Fully automated quantification of retinal imaging biomarkers such as HRF is both reliable and merited as HRF are indicators of retinal pigment epithelium dysmorphia, a central pathogenetic mechanism in GA. Identifying disease markers using the combination of FAF and SD-OCT is of high prognostic value and facilitates individualized patient management in a clinical setting.

Personalized treatment supported by automated quantitative fluid analysis in active neovascular age-related macular degeneration (nAMD)-a phase III, prospective, multicentre, randomized study: design and methods

INTRODUCTION

In neovascular age-related macular degeneration (nAMD) the exact amount of fluid and its location on optical coherence tomography (OCT) have been defined as crucial biomarkers for disease activity and therapeutic decisions. Yet in the absence of quantitative evaluation tools, real-world care outcomes are disappointing. Artificial intelligence (AI) offers a practical option for clinicians to enhance point-of-care management by analysing OCT volumes in a short time. In this protocol we present the prospective implementation of an AI-algorithm providing automated real-time fluid quantifications in a clinical real-world setting.

METHODS

This is a prospective, multicentre, randomized (1:1) and double masked phase III clinical trial. Two-hundred-ninety patients with active nAMD will be randomized between a study arm using AI-supported fluid quantifications and another arm using conventional qualitative assessments, i.e. state-of-the-art disease management. The primary outcome is defined as the mean number of injections over 1 year. Change in BCVA is defined as a secondary outcome.

DISCUSSION

Automated measurement of fluid volumes in all retinal compartments such as intraretinal fluid (IRF), and subretinal fluid (SRF) will serve as an objective tool for clinical investigators on which to base retreatment decisions. Compared to qualitative fluid assessment, retreatment decisions will be plausible and less prone to error or large variability. The underlying hypothesis is that fluid should be treated, while residual persistent or stable amounts of fluid may not benefit from further therapy. Reducing injection numbers without diminishing the visual benefit will increase overall patient safety and relieve the burden for healthcare providers.

TRIAL-REGISTRATION

EudraCT-Number: 2019-003133-42.

Robust Fovea Detection in Retinal OCT Imaging using Deep Learning

The fovea centralis is an essential landmark in the retina where the photoreceptor layer is entirely composed of cones responsible for sharp, central vision. The localization of this anatomical landmark in optical coherence tomography (OCT) volumes is important for assessing visual function correlates and treatment guidance in macular disease. In this study, the "PRE U-net" is introduced as a novel approach for a fully automated fovea centralis detection, addressing the localization as a pixel-wise regression task. 2D B-scans are sampled from each image volume and are concatenated with spatial location information to train the deep network. A total of 5586 OCT volumes from 1,541 eyes was used to train, validate and test the deep learning method. The test data is comprised of healthy subjects and patients affected by neovascular age-related macular degeneration (nAMD), diabetic macula edema (DME) and macular edema from retinal vein occlusion (RVO), covering the three major retinal diseases responsible for blindness. Our experiments demonstrate that the PRE U-net significantly outperforms state-of-the-art methods and improves the robustness of automated localization, which is of value for clinical practice.

Segmentation of Bruch's Membrane in retinal OCT with AMD using anatomical priors and uncertainty quantification

Bruch's membrane (BM) segmentation on optical coherence tomography (OCT) is a pivotal step for the diagnosis and follow-up of age-related macular degeneration (AMD), one of the leading causes of blindness in the developed world. Automated BM segmentation methods exist, but they usually do not account for the anatomical coherence of the results, neither provide feedback on the confidence of the prediction. These factors limit the applicability of these systems in real-world scenarios. With this in mind, we propose an end-to-end deep learning method for automated BM segmentation in AMD patients. An Attention U-Net is trained to output a probability density function of the BM position, while taking into account the natural curvature of the surface. Besides the surface position, the method also estimates an A-scan wise uncertainty measure of the segmentation output. Subsequently, the A-scans with high uncertainty are interpolated using thin plate splines (TPS). We tested our method with ablation studies on an internal dataset with 138 patients covering all three AMD stages, and achieved a mean absolute localization error of 4.10 μm. In addition, the proposed segmentation method was compared against the state-of-the-art methods and showed a superior performance on an external publicly available dataset from a different patient cohort and OCT device, demonstrating strong generalization ability.

Topographic Distribution and Progression of Soft Drusen Volume in Age-Related Macular Degeneration Implicate Neurobiology of Fovea

Purpose

To refine estimates of macular soft drusen abundance in eyes with age-related macular degeneration (AMD) and evaluate hypotheses about drusen biogenesis, we investigated topographic distribution and growth rates of drusen by optical coherence tomography (OCT). We compared results to retinal features with similar topographies (cone density and macular pigment) in healthy eyes.

Methods

In a prospective study, distribution and growth rates of soft drusen in eyes with AMD were identified by human observers in OCT volumes and analyzed with computer-assistance. Published histologic data for macular cone densities (n = 12 eyes) and in vivo macular pigment optical density (MPOD) measurements in older adults with unremarkable maculae (n = 31; 62 paired eyes, averaged) were revisited. All values were normalized to Early Treatment Diabetic Retinopathy Study (ETDRS) subfield areas.

Results

Sixty-two eyes of 44 patients were imaged for periods up to 78 months. Soft drusen volume per unit volume at baseline is 24.6-fold and 2.3-fold higher in the central ETDRS subfield than in outer and inner rings, respectively, and grows most prominently there. Corresponding ratios (central versus inner and central versus outer) for cone density in donor eyes is 13.3-fold and 5.1-fold and for MPOD, 24.6 and 23.9-fold, and 3.6 and 3.6-fold.

Conclusions

Normalized soft drusen volume in AMD eyes as assessed by OCT is ≥ 20-fold higher in central ETDRS subfields than in outer rings, paralleling MPOD distribution in healthy eyes. Data on drusen volume support this metric for AMD risk assessment and clinical trial outcome measure. Alignment of different data modalities support the ETDRS grid for standardizing retinal topography in mechanistic studies of drusen biogenesis.

Quantification of Fluid Resolution and Visual Acuity Gain in Patients With Diabetic Macular Edema Using Deep Learning: A Post Hoc Analysis of a Randomized Clinical Trial

Importance

Large amounts of optical coherence tomographic (OCT) data of diabetic macular edema (DME) are acquired, but many morphologic features have yet to be identified and quantified.

Objective

To examine the volumetric change of intraretinal fluid (IRF) and subretinal fluid (SRF) in DME during anti-vascular endothelial growth factor treatment using deep learning algorithms.

Design, Setting, and Participants

This post hoc analysis of a randomized clinical trial, the Diabetic Retinopathy Clinical Research Network (protocol T), assessed 6945 spectral-domain OCT volume scans of 570 eyes from 570 study participants with DME. The original trial was performed from August 21, 2012, to October 18, 2018. This analysis was performed from December 7, 2017, to January 15, 2020.

Interventions

Participants were treated according to a predefined, standardized protocol with aflibercept, ranibizumab, or bevacizumab with or without deferred laser.

Main Outcomes and Measures

The association of treatment with IRF and SRF volumes and best-corrected visual acuity (BCVA) during 12 months using deep learning algorithms.

Results

Among the 570 study participants (302 [53%] male; 369 [65%] white; mean [SD] age, 43.4 [12.6] years), the mean fluid volumes in the central 3 mm were 448.6 nL (95% CI, 412.3-485.0 nL) of IRF and 36.9 nL (95% CI, 27.0-46.7 nL) of SRF at baseline and 161.2 nL (95% CI, 135.1-187.4 nL) of IRF and 4.4 nL (95% CI, 1.7-7.1 nL) of SRF at 12 months. The presence of SRF at baseline was associated with a worse baseline BCVA Early Treatment Diabetic Retinopathy Study (ETDRS) score of 63.2 (95% CI, 60.2-66.1) (approximate Snellen equivalent of 20/63 [95% CI, 20/50-20/63]) in eyes with SRF vs 66.9 (95% CI, 65.7-68.1) (approximate Snellen equivalent, 20/50 [95% CI, 20/40-20/50]) without SRF (P < .001) and a greater gain in ETDRS score (0.5; 95% CI, 0.3-0.8) every 4 weeks during follow-up in eyes with SRF at baseline vs 0.4 (95% CI, 0.3-0.5) in eyes without SRF at baseline (P = .02) when adjusted for baseline BCVA. Aflibercept was associated with greater reduction of IRF volume compared with bevacizumab after the first injection (difference, 79.8 nL; 95% CI, 5.3-162.5 nL; P < .001) and every 4 weeks thereafter (difference, 10.4 nL; 95% CI, 0.7-20.0 nL; P = .004). Ranibizumab was associated with a greater reduction of IRF after the first injection compared with bevacizumab (difference, 75.2 nL; 95% CI, 1.4-154.7 nL; P < .001).

Conclusions and Relevance

Automated segmentation of fluid in DME revealed that the presence of SRF was associated with lower baseline BCVA but with good response to anti-vascular endothelial growth factor therapy. These automated spectral-domain OCT analyses may be used clinically to assess anatomical change during therapy.

Trial Registration

ClinicalTrials.gov Identifier: NCT01627249.

Characterization of Drusen and Hyperreflective Foci as Biomarkers for Disease Progression in Age-Related Macular Degeneration Using Artificial Intelligence in Optical Coherence Tomography

Importance

The morphologic changes and their pathognomonic distribution in progressing age-related macular degeneration (AMD) are not well understood.

Objectives

To characterize the pathognomonic distribution and time course of morphologic patterns in AMD and to quantify changes distinctive for progression to macular neovascularization (MNV) and macular atrophy (MA).

Design, Setting, and Participants

This cohort study included optical coherence tomography (OCT) volumes from study participants with early or intermediate AMD in the fellow eye in the HARBOR (A Study of Ranibizumab Administered Monthly or on an As-needed Basis in Patients With Subfoveal Neovascular Age-Related Macular Degeneration) trial. Patients underwent imaging monthly for 2 years (July 1, 2009, to August 31, 2012) following a standardized protocol. Data analysis was performed from June 1, 2018, to January 21, 2020.

Main Outcomes and Measures

To obtain topographic correspondence between patients and over time, all scans were mapped into a joint reference frame. The time of progression to MNV and MA was established, and drusen volumes and hyperreflective foci (HRF) volumes were automatically segmented in 3 dimensions using validated artificial intelligence algorithms. Topographically resolved population means of these markers were constructed by averaging quantified drusen and HRF maps in the patient subgroups.

Results

Of 1097 patients enrolled in HARBOR, 518 (mean [SD] age, 78.1 [8.2] years; 309 [59.7%] female) had early or intermediate AMD in the fellow eye at baseline. During the 24-month follow-up period, 135 (26%) eyes developed MNV, 50 eyes (10%) developed MA, and 333 (64%) eyes did not progress to advanced AMD. Drusen and HRF had distinct topographic patterns. Mean drusen thickness at the fovea was 29.6 μm (95% CI, 20.2-39.0 μm) for eyes progressing to MNV, 17.2 μm (95% CI, 9.8-24.6 μm) for eyes progressing to MA, and 17.1 μm (95% CI, 12.5-21.7 μm) for eyes without disease progression. At 0.5-mm eccentricity, mean drusen thickness was 25.8 μm (95% CI, 19.1-32.5 μm) for eyes progressing to MNV, 21.7 μm (95% CI, 14.6-28.8 μm) for eyes progressing to MA, and 14.4 μm (95% CI, 11.2-17.6 μm) for eyes without disease progression. The mean HRF thickness at the foveal center was 0.072 μm (95% CI, 0-0.152 μm) for eyes progressing to MNV, 0.059 μm (95% CI, 0-0.126 μm) for eyes progressing to MA, and 0.044 μm (95% CI, 0.007-0.081) for eyes without disease progression. At 0.5-mm eccentricity, the largest mean HRF thickness was seen in eyes progressing to MA (0.227 μm; 95% CI, 0.104-0.349 μm) followed by eyes progressing to MNV (0.161 μm; 95% CI, 0.101-0.221 μm) and eyes without disease progression (0.085 μm; 95% CI, 0.058-0.112 μm).

Conclusions and Relevance

In this study, drusen and HRF represented imaging biomarkers of disease progression in AMD, demonstrating distinct topographic patterns over time that differed between eyes progressing to MNV, eyes progressing to MA, or eyes without disease progression. Automated localization and precise quantification of these factors may help to develop reliable methods of predicting future disease progression.

End-to-End Deep Learning Model for Predicting Treatment Requirements in Neovascular AMD From Longitudinal Retinal OCT Imaging

Neovascular age-related macular degeneration (nAMD) is nowadays successfully treated with anti-VEGF substances, but inter-individual treatment requirements are vastly heterogeneous and currently poorly plannable resulting in suboptimal treatment frequency. Optical coherence tomography (OCT) with its 3D high-resolution imaging serves as a companion diagnostic to anti-VEGF therapy. This creates a need for building predictive models using automated image analysis of OCT scans acquired during the treatment initiation phase. We propose such a model based on deep learning (DL) architecture, comprised of a densely connected neural network (DenseNet) and a recurrent neural network (RNN), trainable end-to-end. The method starts by sampling several 2D-images from an OCT volume to obtain a lower-dimensional OCT representation. At the core of the predictive model, the DenseNet learns useful retinal spatial features while the RNN integrates information from different time points. The introduced model was evaluated on the prediction of anti-VEGF treatment requirements in nAMD patients treated under a pro-re-nata (PRN) regimen. The DL model was trained on 281 patients and evaluated on a hold-out test set of 69 patient. The predictive model achieved a concordance index of 0.7 in regressing the number of received treatments, while in a classification task it obtained an 0.85 (0.81) AUC in detecting the patients with low (high) treatment requirements. The proposed model outperformed previous machine learning strategies that relied on a set of spatio-temporal image features, showing that the proposed DL architecture successfully learned to extract the relevant spatio-temporal patterns directly from raw longitudinal OCT images.

Automated quantification of macular fluid in retinal diseases and their response to anti-VEGF therapy

AIM

To objectively assess disease activity and treatment response in patients with retinal vein occlusion (RVO), neovascular age-related macular degeneration (nAMD) and centre-involved diabetic macular oedema (DME), using artificial intelligence-based fluid quantification.

METHODS

Posthoc analysis of 2311 patients (11 151 spectral-domain optical coherence tomography volumes) from five clinical, multicentre trials, who received a flexible antivascular endothelial growth factor (anti-VEGF) therapy over a 12-month period. Fluid volumes were measured with a deep learning algorithm at baseline/months 1, 2, 3 and 12, for three concentric circles with diameters of 1, 3 and 6 mm (fovea, paracentral ring and pericentral ring), as well as four sectors surrounding the fovea (superior, nasal, inferior and temporal).

RESULTS

In each disease, at every timepoint, most intraretinal fluid (IRF) per square millimetre was present at the fovea, followed by the paracentral ring and pericentral ring (p<0.0001). While this was also the case for subretinal fluid (SRF) in RVO/DME (p<0.0001), patients with nAMD showed more SRF in the paracentral ring than at the fovea up to month 3 (p<0.0001). Between sectors, patients with RVO/DME showed the highest IRF volumes temporally (p<0.001/p<0.0001). In each disease, more SRF was consistently found inferiorly than superiorly (p<0.02). At month 1/12, we measured the following median reductions of initial fluid volumes. For IRF: RVO, 95.9%/97.7%; nAMD, 91.3%/92.8%; DME, 37.3%/69.9%. For SRF: RVO, 94.7%/97.5%; nAMD, 98.4%/99.8%; DME, 86.3%/97.5%.

CONCLUSION

Fully automated localisation and quantification of IRF/SRF over time shed light on the fluid dynamics in each disease. There is a specific anatomical response of IRF/SRF to anti-VEGF therapy in all diseases studied.

Role of Deep Learning-Quantified Hyperreflective Foci for the Prediction of Geographic Atrophy Progression

PURPOSE

To quantitatively measure hyperreflective foci (HRF) during the progression of geographic atrophy (GA) secondary to age-related macular degeneration (AMD) using deep learning (DL) and investigate the association with local and global growth of GA.

METHODS

Eyes with GA were prospectively included. Spectral-domain optical coherence tomography (SDOCT) and fundus autofluorescence images were acquired every 6 months. A 500-μm-wide junctional zone adjacent to the GA border was delineated and HRF were quantified using a validated DL algorithm. HRF concentrations in progressing and nonprogressing areas, as well as correlations between HRF quantifications and global and local GA progression, were assessed.

RESULTS

A total of 491 SDOCT volumes from 87 eyes of 54 patients were assessed with a median follow-up of 28 months. Two-thirds of HRF were localized within a millimeter adjacent to the GA border. HRF concentration was positively correlated with GA progression in unifocal and multifocal GA (all P < .001) and de novo GA development (P = .037). Local progression speed correlated positively with local increase of HRF (P value range <.001-.004). Global progression speed, however, did not correlate with HRF concentrations (P > .05). Changes in HRF over time did not have an impact on the growth in GA (P > .05).

CONCLUSION

Advanced artificial intelligence (AI) methods in high-resolution retinal imaging allows to identify, localize, and quantify biomarkers such as HRF. Increased HRF concentrations in the junctional zone and future macular atrophy may represent progressive migration and loss of retinal pigment epithelium. AI-based biomarker monitoring may pave the way into the era of individualized risk assessment and objective decision-making processes. NOTE: Publication of this article is sponsored by the American Ophthalmological Society.

Correction to "Exploiting Epistemic Uncertainty of Anatomy Segmentation for Anomaly Detection in Retinal OCT"

The authors of "Exploiting Epistemic Uncertainty of Anatomy Segmentation for Anomaly Detection in Retinal OCT" which appeared in the January 2020 issue of this journal [1] would like to provide an updated Fig. 3 because there was an error in the published version. The output of the last convolutional layers says "2" in the number of channels but it should be "11" (10 retinal layer and the background).

Exploiting Epistemic Uncertainty of Anatomy Segmentation for Anomaly Detection in Retinal OCT

Diagnosis and treatment guidance are aided by detecting relevant biomarkers in medical images. Although supervised deep learning can perform accurate segmentation of pathological areas, it is limited by requiring a priori definitions of these regions, large-scale annotations, and a representative patient cohort in the training set. In contrast, anomaly detection is not limited to specific definitions of pathologies and allows for training on healthy samples without annotation. Anomalous regions can then serve as candidates for biomarker discovery. Knowledge about normal anatomical structure brings implicit information for detecting anomalies. We propose to take advantage of this property using Bayesian deep learning, based on the assumption that epistemic uncertainties will correlate with anatomical deviations from a normal training set. A Bayesian U-Net is trained on a well-defined healthy environment using weak labels of healthy anatomy produced by existing methods. At test time, we capture epistemic uncertainty estimates of our model using Monte Carlo dropout. A novel post-processing technique is then applied to exploit these estimates and transfer their layered appearance to smooth blob-shaped segmentations of the anomalies. We experimentally validated this approach in retinal optical coherence tomography (OCT) images, using weak labels of retinal layers. Our method achieved a Dice index of 0.789 in an independent anomaly test set of age-related macular degeneration (AMD) cases. The resulting segmentations allowed very high accuracy for separating healthy and diseased cases with late wet AMD, dry geographic atrophy (GA), diabetic macular edema (DME) and retinal vein occlusion (RVO). Finally, we qualitatively observed that our approach can also detect other deviations in normal scans such as cut edge artifacts.

RETOUCH: The Retinal OCT Fluid Detection and Segmentation Benchmark and Challenge

Retinal swelling due to the accumulation of fluid is associated with the most vision-threatening retinal diseases. Optical coherence tomography (OCT) is the current standard of care in assessing the presence and quantity of retinal fluid and image-guided treatment management. Deep learning methods have made their impact across medical imaging, and many retinal OCT analysis methods have been proposed. However, it is currently not clear how successful they are in interpreting the retinal fluid on OCT, which is due to the lack of standardized benchmarks. To address this, we organized a challenge RETOUCH in conjunction with MICCAI 2017, with eight teams participating. The challenge consisted of two tasks: fluid detection and fluid segmentation. It featured for the first time: all three retinal fluid types, with annotated images provided by two clinical centers, which were acquired with the three most common OCT device vendors from patients with two different retinal diseases. The analysis revealed that in the detection task, the performance on the automated fluid detection was within the inter-grader variability. However, in the segmentation task, fusing the automated methods produced segmentations that were superior to all individual methods, indicating the need for further improvements in the segmentation performance.

Unsupervised Identification of Disease Marker Candidates in Retinal OCT Imaging Data

The identification and quantification of markers in medical images is critical for diagnosis, prognosis, and disease management. Supervised machine learning enables the detection and exploitation of findings that are known a priori after annotation of training examples by experts. However, supervision does not scale well, due to the amount of necessary training examples, and the limitation of the marker vocabulary to known entities. In this proof-of-concept study, we propose unsupervised identification of anomalies as candidates for markers in retinal optical coherence tomography (OCT) imaging data without a constraint to a priori definitions. We identify and categorize marker candidates occurring frequently in the data and demonstrate that these markers show a predictive value in the task of detecting disease. A careful qualitative analysis of the identified data driven markers reveals how their quantifiable occurrence aligns with our current understanding of disease course, in early- and late age-related macular degeneration (AMD) patients. A multi-scale deep denoising autoencoder is trained on healthy images, and a one-class support vector machine identifies anomalies in new data. Clustering in the anomalies identifies stable categories. Using these markers to classify healthy-, early AMD- and late AMD cases yields an accuracy of 81.40%. In a second binary classification experiment on a publicly available data set (healthy versus intermediate AMD), the model achieves an area under the ROC curve of 0.944.

Prediction of Individual Disease Conversion in Early AMD Using Artificial Intelligence

Purpose

While millions of individuals show early age-related macular degeneration (AMD) signs, yet have excellent vision, the risk of progression to advanced AMD with legal blindness is highly variable. We suggest means of artificial intelligence to individually predict AMD progression.

Methods

In eyes with intermediate AMD, progression to the neovascular type with choroidal neovascularization (CNV) or the dry type with geographic atrophy (GA) was diagnosed based on standardized monthly optical coherence tomography (OCT) images by independent graders. We obtained automated volumetric segmentation of outer neurosensory layers and retinal pigment epithelium, drusen, and hyperreflective foci by spectral domain-OCT image analysis. Using imaging, demographic, and genetic input features, we developed and validated a machine learning-based predictive model assessing the risk of conversion to advanced AMD.

Results

Of a total of 495 eyes, 159 eyes (32%) had converted to advanced AMD within 2 years, 114 eyes progressed to CNV, and 45 to GA. Our predictive model differentiated converting versus nonconverting eyes with a performance of 0.68 and 0.80 for CNV and GA, respectively. The most critical quantitative features for progression were outer retinal thickness, hyperreflective foci, and drusen area. The features for conversion showed pathognomonic patterns that were distinctly different for the neovascular and the atrophic pathways. Predictive hallmarks for CNV were mostly drusen-centric, while GA markers were associated with neurosensory retina and age.

Conclusions

Artificial intelligence with automated analysis of imaging biomarkers allows personalized prediction of AMD progression. Moreover, pathways of progression may be specific in respect to the neovascular/atrophic type.

Why rankings of biomedical image analysis competitions should be interpreted with care

International challenges have become the standard for validation of biomedical image analysis methods. Given their scientific impact, it is surprising that a critical analysis of common practices related to the organization of challenges has not yet been performed. In this paper, we present a comprehensive analysis of biomedical image analysis challenges conducted up to now. We demonstrate the importance of challenges and show that the lack of quality control has critical consequences. First, reproducibility and interpretation of the results is often hampered as only a fraction of relevant information is typically provided. Second, the rank of an algorithm is generally not robust to a number of variables such as the test data used for validation, the ranking scheme applied and the observers that make the reference annotations. To overcome these problems, we recommend best practice guidelines and define open research questions to be addressed in the future.

Neuroretinal atrophy following resolution of macular oedema in retinal vein occlusion

BACKGROUND/AIMS

To characterise neuroretinal atrophy in retinal vein occlusion (RVO).

METHODS

We included patients with central/branch RVO (CRVO=196, BRVO=107) who received ranibizumab according to a standardised protocol for 6 months. Retinal atrophy was defined as the presence of an area of retinal thickness (RT) <260 µm outside the foveal centre. Moreover, the thickness of three distinct retinal layer compartments was computed as follows: (1) retinal nerve fibre layer to ganglion cell layer, (2) inner plexiform layer (IPL) to outer nuclear layer (ONL) and (3) inner segment/outer segment junction to retinal pigment epithelium. To characterise atrophy further, we assessed perfusion status on fluorescein angiography and best-corrected visual acuity (BCVA), and compared these between eyes with/without atrophy.

RESULTS

23 patients with CRVO and 11 patients with BRVO demonstrated retinal atrophy, presenting as sharply demarcated retinal thinning confined to a macular quadrant. The mean RT in the atrophic quadrant at month 6 was 249±26 µm (CRVO) and 244±29 µm (BRVO). Individual layer analysis revealed pronounced thinning in the IPL to ONL compartment. Change in BCVA at 6 months was similar between the groups (BRVO, +15 vs +18 letters; CRVO, +14 vs +18 letters).

CONCLUSIONS

In this exploratory analysis, we describe the characteristics of neuroretinal atrophy in RVO eyes with resolved macular oedema after ranibizumab therapy. Our analysis shows significant, predominantly retinal thinning in the IPL to ONL compartment in focal macular areas in 11% of patients with RVO. Eyes with retinal atrophy did not show poorer BCVA outcomes.

Spatial Correspondence Between Intraretinal Fluid, Subretinal Fluid, and Pigment Epithelial Detachment in Neovascular Age-Related Macular Degeneration

Purpose

To identify the spatial distribution of exudative features of choroidal neovascularization in neovascular age-related macular degeneration (nAMD) based on the localization of intraretinal cystoid fluid (IRC), subretinal fluid (SRF), and pigment-epithelial detachment (PED).

Methods

This retrospective cross-sectional study included spectral-domain optical coherence tomography volume scans (6 × 6 mm) of 1341 patients with treatment-naïve nAMD. IRC, SRF, and PED were detected on a per-voxel basis using fully automated segmentation algorithms. Two subsets of 37 volumes each were manually segmented to validate the automated results. The spatial correspondence of components was quantified by computing proportions of IRC-, SRF-, or PED-presenting A-scans simultaneously affected by the respective other pathomorphologic components on a per-patient basis. The median across the population is reported. Odds ratios between pairs of lesions were calculated and tested for significance pixel wise.

Results

Automated image segmentation was successful in 1182 optical coherence tomography volumes, yielding more than 61 million A-scans for analysis. Overall, 81% of eyes showed IRC, 95% showed SRF, and 92% showed PED. IRC-presenting A-scans also showed SRF in a median 2.5%, PED in 32.9%. Of the SRF-presenting A-scans, 0.3% demonstrated IRC, 1.4% PED. Of the PED-presenting A-scans, 5.2% contained IRC, 2.0% SRF. Similar patterns were observed in the manually segmented subsets and via pixel-wise odds ratio analysis.

Conclusions

Automated analyses of large-scale datasets in a cross-sectional study of 1182 patients with active treatment-naïve nAMD demonstrated low spatial correlation of SRF with IRC and PED in contrast to increased colocalization of IRC and PED. These morphological associations may contribute to our understanding of functional deficits in nAMD.

Supervised learning and dimension reduction techniques for quantification of retinal fluid in optical coherence tomography images

PurposeThe purpose of the present study is to develop fast automated quantification of retinal fluid in optical coherence tomography (OCT) image sets.MethodsWe developed an image analysis pipeline tailored towards OCT images that consists of five steps for binary retinal fluid segmentation. The method is based on feature extraction, pre-segmention, dimension reduction procedures, and supervised learning tools.ResultsFluid identification using our pipeline was tested on two separate patient groups: one associated to neovascular age-related macular degeneration, the other showing diabetic macular edema. For training and evaluation purposes, retinal fluid was annotated manually in each cross-section by human expert graders of the Vienna Reading Center. Compared with the manual annotations, our pipeline yields good quantification, visually and in numbers.ConclusionsBy demonstrating good automated retinal fluid quantification, our pipeline appears useful to expert graders within their current grading processes. Owing to dimension reduction, the actual learning part is fast and requires only few training samples. Hence, it is well-suited for integration into actual manufacturer's devices, further improving segmentation by its use in daily clinical life.

Machine Learning of the Progression of Intermediate Age-Related Macular Degeneration Based on OCT Imaging

Purpose

To develop a data-driven interpretable predictive model of incoming drusen regression as a sign of disease activity and identify optical coherence tomography (OCT) biomarkers associated with its risk in intermediate age-related macular degeneration (AMD).

Methods

Patients with AMD were observed every 3 months, using Spectralis OCT imaging, for a minimum duration of 12 months and up to a period of 60 months. Segmentation of drusen and the overlying layers was obtained using a graph-theoretic method, and the hyperreflective foci were segmented using a voxel classification method. Automated image analysis steps were then applied to identify and characterize individual drusen at baseline, and their development was monitored at every follow-up visit. Finally, a machine learning method based on a sparse Cox proportional hazard regression was developed to estimate a risk score and predict the incoming regression of individual drusen.

Results

The predictive model was trained and evaluated on a longitudinal dataset of 61 eyes from 38 patients using cross-validation. The mean follow-up time was 37.8 ± 13.8 months. A total of 944 drusen were identified at baseline, out of which 249 (26%) regressed during follow-up. The prediction performance was evaluated as area under the curve (AUC) for different time periods. Prediction within the first 2 years achieved an AUC of 0.75.

Conclusions

The predictive model proposed in this study represents a promising step toward image-guided prediction of AMD progression. Machine learning is expected to accelerate and contribute to the development of new therapeutics that delay the progression of AMD.

Prediction of Anti-VEGF Treatment Requirements in Neovascular AMD Using a Machine Learning Approach

Purpose

The purpose of this study was to predict low and high anti-VEGF injection requirements during a pro re nata (PRN) treatment, based on sets of optical coherence tomography (OCT) images acquired during the initiation phase in neovascular AMD.

Methods

Two-year clinical trial data of subjects receiving PRN ranibizumab according to protocol specified criteria in the HARBOR study after three initial monthly injections were included. OCT images were analyzed at baseline, month 1, and month 2. Quantitative spatio-temporal features computed from automated segmentation of retinal layers and fluid-filled regions were used to describe the macular microstructure. In addition, best-corrected visual acuity and demographic characteristics were included. Patients were grouped into low and high treatment categories based on first and third quartile, respectively. Random forest classification was used to learn and predict treatment categories and was evaluated with cross-validation.

Results

Of 317 evaluable subjects, 71 patients presented low (≤5), 176 medium, and 70 high (≥16) injection requirements during the PRN maintenance phase from month 3 to month 23. Classification of low and high treatment requirement subgroups demonstrated an area under the receiver operating characteristic curve of 0.7 and 0.77, respectively. The most relevant feature for prediction was subretinal fluid volume in the central 3 mm, with the highest predictive values at month 2.

Conclusions

We proposed and evaluated a machine learning methodology to predict anti-VEGF treatment needs from OCT scans taken during treatment initiation. The results of this pilot study are an important step toward image-guided prediction of treatment intervals in the management of neovascular AMD.

Automated Segmentability Index for Layer Segmentation of Macular SD-OCT Images

PURPOSE

To automatically identify which spectral-domain optical coherence tomography (SD-OCT) scans will provide reliable automated layer segmentations for more accurate layer thickness analyses in population studies.

METHODS

Six hundred ninety macular SD-OCT image volumes (6.0 × 6.0 × 2.3 mm³) were obtained from one eyes of 690 subjects (74.6 ± 9.7 [mean ± SD] years, 37.8% of males) randomly selected from the population-based Rotterdam Study. The dataset consisted of 420 OCT volumes with successful automated retinal nerve fiber layer (RNFL) segmentations obtained from our previously reported graph-based segmentation method and 270 volumes with failed segmentations. To evaluate the reliability of the layer segmentations, we have developed a new metric, segmentability index SI, which is obtained from a random forest regressor based on 12 features using OCT voxel intensities, edge-based costs, and on-surface costs. The SI was compared with well-known quality indices, quality index (QI), and maximum tissue contrast index (mTCI), using receiver operating characteristic (ROC) analysis.

RESULTS

The 95% confidence interval (CI) and the area under the curve (AUC) for the QI are 0.621 to 0.805 with AUC 0.713, for the mTCI 0.673 to 0.838 with AUC 0.756, and for the SI 0.784 to 0.920 with AUC 0.852. The SI AUC is significantly larger than either the QI or mTCI AUC (P < 0.01).

CONCLUSIONS

The segmentability index SI is well suited to identify SD-OCT scans for which successful automated intraretinal layer segmentations can be expected.

TRANSLATIONAL RELEVANCE

Interpreting the quantification of SD-OCT images requires the underlying segmentation to be reliable, but standard SD-OCT quality metrics do not predict which segmentations are reliable and which are not. The segmentability index SI presented in this study does allow reliable segmentations to be identified, which is important for more accurate layer thickness analyses in research and population studies.