Automated Large-scale Prediction of Exudative And Progression Using Machine-read Oct Biomarkers

Akos Rudas, Jeffrey N. Chiang, Giulia Corradetti, Nadav Rakocz, Oren Avram, Eran Halperin, Srinivas R. Sadda

Age-related Macular Degeneration (AMD) is a major cause of irreversible vision loss in individuals over 55 years old in the United States. One of the late-stage manifestations of AMD, and a major cause of vision loss, is the development of exudative macular neovascularization (MNV). Optical Coherence Tomography (OCT) is the gold standard to identify fluid at different levels within the retina. The presence of fluid is considered the hallmark to define the presence of disease activity. Anti-vascular growth factor (anti-VEGF) injections can be used to treat exudative MNV. However, given the limitations of anti-VEGF treatment, as burdensome need for frequent visits and repeated injections to sustain efficacy, limited durability of the treatment, poor or no response, there is a great interest in detecting early biomarkers associated with a higher risk for AMD progression to exudative forms in order to optimize the design of early intervention clinical trials. The annotation of structural biomarkers on optical coherence tomography (OCT) B-scans is a laborious, complex and time-consuming process, and discrepancies between human graders can introduce variability into this assessment.

Age-related Macular Degeneration (AMD) represents the leading cause of irreversible blindness in subjects older than 55 years of age in developed countries [1]. As the population ages and life expectancy increases, the incidence of the disease is projected to rise [2]. The late stage of the disease is characterized by the presence of geographic atrophy (GA), macular atrophy (MA) or macular neovascularization (MNV) [3–5].

Materials and methods
Study design and dataset

The study was conducted in compliance with the Declaration of Helsinki and approved by the UCLA Institutional Review Board (IRB, Ocular Imaging Study; Doheny–UCLA Eye Centers).

The dataset consisted of 14,615 OCT volumes collected from 4,182 patients at affiliated Ophthalmology clinics during 2018 and corresponding electronic health record data for these visits including demographics, AMD status, and comorbidities (see Table 5). OCT volumes were obtained by the Spectralis OCT device (19 B-scans, 20x20 degree centered on the fovea). A single volume for each (exam date, patient, eye) triplet was included in the study. Volumes collected during the same encounter and corresponding to the same eye were aggregated, selecting the maximum measured value for each biomarker on that date. It should be noted that since the dataset in this study was selected from a specific time frame, progression-wise the data is right-censored. Examination in a survival analysis framework is in the scope of future work.

Machine-read OCT features were evaluated for their clinical utility relative to currently known risk factors contained within the electronic health record using a predictive modeling framework. These features were evaluated in their ability to predict conversion to exudative AMD as well as diagnosis of current exudative AMD.

Our study also has a few strengths. First, the machine learning algorithms have been trained and tested on a large cohort. We have performed a large-scale automatic validation of these previously established biomarkers, validating not only the biomarkers, but their automatic identification as well. Furthermore, we have provided evidence that automatic detection of structural OCT B-scan biomarkers using machine learning can be of value in predicting exudative AMD. The algorithm has the ability to provide automated annotation of these biomarkers on OCT volumes with high precision and feasibility, avoiding the laborious manual inspection or annotation of all the OCT B-scans. Also, considering the challenges associated with implementing and deploying separate models for different time horizons, the 2-year model was separately evaluated on different time frames. Since no significant drop in performance was observed, it is reasonable to assume that the model can successfully utilize the provided timedelta feature. Thus, we determined that it is sufficient to deploy a single model across different time frames.

In conclusion, we demonstrate on a large dataset that a machine learning algorithm can automatically annotate OCT volumes with high-risk structural OCT B-scan biomarkers of AMD progression with high accuracy. These annotations can be used to predict conversion to exudative AMD in eyes with nonexudative AMD with good performance, providing an impactful example of how machine learning has the ability to enhance patient care.

Citation: Rudas A, Chiang JN, Corradetti G, Rakocz N, Avram O, Halperin E, et al. (2023) Automated large-scale prediction of exudative AMD progression using machine-read OCT biomarkers. PLOS Digit Health 2(2): e0000106.

Editor: Nan Liu, Duke-NUS Medical School, SINGAPORE

Received: August 12, 2022; Accepted: January 14, 2023; Published: February 15, 2023

Copyright: © 2023 Rudas et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: The data are not publicly available due to institutional data use policy and concerns about patient privacy. The data set consists of 14,615 OCT volumes and corresponding electronic health record data collected from 4,182 patients in 2018 at Doheny UCLA Eye Centers. In order to apply for data access, please visit or reach out to

Funding: The authors received no specific funding for this work.

Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: Eran Halperin has received consulting fees from United Health Group. SriniVas Sadda has received consulting fees from Amgen, Allergan, Genentech-Roche, Oxurion, Novartis, Regeneron, Iveric, 4DMT, Centervue, Heidelberg, Optos, Carl Zeiss Meditec, Nidek, and Topcon.

Equinix accelerated medical research...The Healthcare Patient Experience & Engagement Summit 2024Healthcare Innovation & Transformation SummitHealthcare CNO SummitHealthcare CMO Summit