Related Experiment Video
Updated: Nov 4, 2025
10:14
Author Spotlight: Ex Vivo OCT-Based Multimodal Imaging of Human Donor Eyes for Research into Age-Related Macular Degeneration
Published on: May 26, 2023
3.8K
Assessing the Clinical Utility of Expanded Macular OCTs Using Machine Learning
Andrew C Lin1,2, Cecilia S Lee1, Marian Blazes1
1Department of Ophthalmology, School of Medicine, University of Washington, Seattle, WA, USA.
Translational Vision Science & Technology
|May 26, 2021
Summary
Expanding optical coherence tomography (OCT) B-scans improves diagnostic accuracy for retinal diseases like AMD, DME, and POAG. Machine learning models show optimal diagnostic performance with specific macular coverage ranges for each condition.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Optical coherence tomography (OCT) is crucial for diagnosing retinal diseases such as age-related macular degeneration (AMD), diabetic macular edema (DME), and primary open-angle glaucoma (POAG).
- Current OCT imaging standards primarily focus on foveal regions, potentially limiting diagnostic performance for certain pathologies.
Purpose of the Study:
- To evaluate the diagnostic performance gains of expanded macular OCT B-scans compared to foveal-only scans for AMD, DME, and POAG.
- To determine the optimal extent of macular coverage for artificial intelligence (AI)-assisted diagnosis in these common retinal diseases.
Main Methods:
- Utilized machine learning techniques, including deep neural networks and random forest ensembles, on a dataset of 630,000 OCT images from patients with AMD, diabetic retinopathy, or POAG.
- Generated area under the receiver operating characteristic (AUROC) and area under the precision recall (AUPR) curves to assess diagnostic performance.
- Analyzed diagnostic accuracy gains as a function of increasing macular OCT B-scan coverage.
Main Results:
- Achieved improved AUROC and AUPR curves when comparing all macular scans (61 B-scans) versus a single central B-scan.
- Identified points of diminishing returns for diagnostic accuracy, with optimal ranges varying by disease: 2.75-4.00 mm (AMD), 4.25-4.50 mm (DME), and 4.50-6.25 mm (POAG).
- Demonstrated statistically significant improvements in diagnostic accuracy for all diseases with >0.25 mm of macular coverage (P < 0.05).
Conclusions:
- Expanded macular coverage in OCT imaging significantly impacts AI-based diagnostic support for AMD, DME, and POAG.
- The required macular area for improved diagnostic accuracy varies, with POAG requiring the largest coverage, followed by DME and AMD.
- These findings support the optimization of OCT imaging protocols to enhance AI decision support for retinal disease management.
