Geographic Atrophy in AMD: Prognostic Factors Based on Long-Term Follow-Up.

INTRODUCTION: The aim of this large-scale long-term retrospective study was to show the enlargement rate (ER) of geographic atrophy (GA) in age-related macular degeneration (AMD), defined as complete retinal pigment epithelium and outer retinal atrophy (cRORA), to find predictors of progression in a clinical routine setting and to compare GA evaluation methods. METHODS: All patients available in our database with follow-up of at least 24 months and cRORA in at least one eye, regardless of neovascular AMD being present, were included. SD-OCT and fundus autofluorescence (FAF) evaluations were performed according to a standardized protocol. The cRORA area ER, the cRORA square root area ER, the FAF GA area, and the condition of the outer retina (inner-/outer-segment [IS/OS] line and external limiting membrane [ELM] disruption scores) were determined. RESULTS: 204 eyes of 129 patients were included. Mean follow-up time was 4.2 ± 2.2 (range 2-10) years. 109 of 204 (53.4%) eyes were classified as MNV-associated GA in AMD (initially or during follow-up); 95 of 204 (46.6%) eyes were classified as pure GA in AMD. The primary lesion was unifocal in 146 (72%) eyes and multifocal in 58 (28%) eyes. A strong correlation was observed between the area of cRORA (SD-OCT) and the FAF GA area (r = 0.924; p < 0.001). Mean ER was 1.44 ± 1.2 mm2/year, mean square root ER 0.29 ± 0.19 mm/year. There was no significant difference in mean ER between eyes without (pure GA) and with intravitreal anti-VEGF injections (MNV-associated GA) (0.30 ± 0.19 mm/year vs. 0.28 ± 0.20 mm/year; p = 0.466). Eyes with multifocal atrophy pattern at baseline had a significantly higher mean ER compared to eyes with unifocal pattern (0.34 ± 0.19 mm/year vs. 0.27 ± 1.19 mm/year; p = 0.008). There were moderate significant correlations between ELM and IS/OS disruption scores and visual acuity at baseline, 5 and 7 years (all r values ca. -0.5; p < 0.001). In multivariate regression analysis, a multifocal cRORA pattern at baseline (p = 0.022) and a smaller baseline lesion size (p = 0.036) were associated with a higher mean ER. CONCLUSION: SD-OCT-evaluated cRORA area might serve as a GA parameter comparable to traditional FAF measurement in clinical routine. The dispersion pattern and baseline lesion size might be predictors of ER, whereas anti-VEGF treatment seems not to be associated with ER.

Optical coherence tomography predictors of progression of non-exudative age-related macular degeneration to advanced atrophic and exudative disease.

PURPOSE: To study the natural history of optical coherence tomography (OCT) imaging-based findings seen in non-exudative age-related macular degeneration (neAMD) and model their relative likelihood in predicting development of incomplete retinal pigment epithelium and outer retinal atrophy (iRORA), complete retinal pigment epithelium and outer retinal atrophy (cRORA), and neovascular AMD (nAMD). METHODS: Retrospective chart review was performed at two academic practices. Patients diagnosed with neAMD for whom yearly OCT scans were obtained for at least 4 consecutive years were included. Baseline demographic, visual acuity, AREDS staging, and OCT data were collected. OCTs were assessed for the presence or absence of eleven features previously individually associated with progression of neAMD, both at baseline, and on all subsequent follow-up scans. Likewise, charts were reviewed to assess visual acuity and staging of NEAMD at all follow-up visits. A multivariate regression analysis was constructed to determine predictors of iRORA, cRORA, and nAMD. RESULTS: A total of 107 eyes of 88 patients were evaluated. Follow-up included yearly OCTs obtained over at least 4 consecutive years follow-up (range: 50-94 months). During the follow-up period, 17 eyes progressed to iRORA while 25 progressed to cRORA and 16 underwent conversion to nAMD. Predictors of conversion to iRORA and cRORA included integrity of the external limiting membrane (p = 0.02), the ellipsoid zone (p = 0.01), and the cone outer segment line (p = 0.003) and the presence of intraretinal hyporeflective spaces (p = 0.009), drusen ooze (p = 0.05), and drusen collapse (p = 0.001). OCT features predictive of conversion to nAMD included outer nuclear layer (ONL) loss (p = 0.01), presence of intraretinal (p = 0.001) and subretinal (p = 0.005) hyporeflective spaces, and drusen collapse (p = 0.003). CONCLUSION: Of these multiple factors predictive of progression of neAMD, the OCT feature most strongly correlated to progression to iRORA/cRORA was drusen collapse, and the feature most predictive of conversion to nAMD was the presence of intraretinal hyporeflective spaces.

Atrophic Lesions Associated with Age-Related Macular Degeneration: High-Resolution versus Standard OCT.

OBJECTIVE: The objective of this study was to determine whether high-resolution OCT (HR-OCT) could enhance the identification and classification of atrophic features in age-related macular degeneration (AMD) compared with standard resolution OCT. DESIGN: Prospective, observational, cross-sectional study. SUBJECTS: The study included 60 eyes from 60 patients > 60 years of age with a diagnosis of AMD. METHODS: The participants underwent volume OCT scanning using HR-OCT and standard resolution OCT devices. Trained graders reviewed and graded the scans, identifying specific regions of interest for subsequent analysis. MAIN OUTCOME MEASURES: The study focused on identifying and classifying complete retinal pigment epithelium (RPE) and outer retinal atrophy (cRORA), incomplete RORA (iRORA), and other nonatrophic AMD features. Additionally, qualitative and quantitative features associated with atrophy were assessed. RESULTS: The agreement among readers for classifying atrophic lesions was substantial to perfect for both HR-OCT (0.88) and standard resolution OCT(0.82). However, HR-OCT showed a higher accuracy in identifying iRORA lesions compared with standard OCT. Qualitative assessment of features demonstrated higher agreement for HR-OCT, particularly in identifying external limiting membrane (ELM) (0.95) and ellipsoid zone (EZ) disruption (0.94). Quantitative measurements of features such as hypertransmission defects, RPE attenuation/disruption, EZ disruption width, and ELM disruption width showed excellent interreader agreement with HR-OCT (> 0.90 for all features) but only moderate agreement with standard OCT (0.51-0.60). CONCLUSIONS: The study results suggest that HR-OCT improves the accuracy and reliability of classifying and quantifying atrophic lesions associated with AMD compared with standard resolution OCT. The quantitative findings in particular may have implications for future research and clinical practice, especially with the availability of therapeutic agents for treating geographic atrophy and the development of commercially available HR-OCT devices. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Improving Interpretability in Machine Diagnosis: Detection of Geographic Atrophy in OCT Scans.

Purpose: Manually identifying geographic atrophy (GA) presence and location on OCT volume scans can be challenging and time consuming. This study developed a deep learning model simultaneously (1) to perform automated detection of GA presence or absence from OCT volume scans and (2) to provide interpretability by demonstrating which regions of which B-scans show GA. Design: Med-XAI-Net, an interpretable deep learning model was developed to detect GA presence or absence from OCT volume scans using only volume scan labels, as well as to interpret the most relevant B-scans and B-scan regions. Participants: One thousand two hundred eighty-four OCT volume scans (each containing 100 B-scans) from 311 participants, including 321 volumes with GA and 963 volumes without GA. Methods: Med-XAI-Net simulates the human diagnostic process by using a region-attention module to locate the most relevant region in each B-scan, followed by an image-attention module to select the most relevant B-scans for classifying GA presence or absence in each OCT volume scan. Med-XAI-Net was trained and tested (80% and 20% participants, respectively) using gold standard volume scan labels from human expert graders. Main Outcome Measures: Accuracy, area under the receiver operating characteristic (ROC) curve, F1 score, sensitivity, and specificity. Results: In the detection of GA presence or absence, Med-XAI-Net obtained superior performance (91.5%, 93.5%, 82.3%, 82.8%, and 94.6% on accuracy, area under the ROC curve, F1 score, sensitivity, and specificity, respectively) to that of 2 other state-of-the-art deep learning methods. The performance of ophthalmologists grading only the 5 B-scans selected by Med-XAI-Net as most relevant (95.7%, 95.4%, 91.2%, and 100%, respectively) was almost identical to that of ophthalmologists grading all volume scans (96.0%, 95.7%, 91.8%, and 100%, respectively). Even grading only 1 region in 1 B-scan, the ophthalmologists demonstrated moderately high performance (89.0%, 87.4%, 77.6%, and 100%, respectively). Conclusions: Despite using ground truth labels during training at the volume scan level only, Med-XAI-Net was effective in locating GA in B-scans and selecting relevant B-scans within each volume scan for GA diagnosis. These results illustrate the strengths of Med-XAI-Net in interpreting which regions and B-scans contribute to GA detection in the volume scan.