Related Experiment Video
Updated: Dec 7, 2025
08:22
Application of Optical Coherence Tomography to a Mouse Model of Retinopathy
Published on: January 12, 2022
4.9K
Development of Deep Learning Models to Predict Best-Corrected Visual Acuity from Optical Coherence Tomography
Michael G Kawczynski1, Thomas Bengtsson1, Jian Dai1
1Genentech, Inc., South San Francisco, CA, USA.
Translational Vision Science & Technology
|September 25, 2020
Summary
Deep learning models accurately predict visual acuity from optical coherence tomography scans in patients with neovascular age-related macular degeneration (nAMD). These models show promise for clinical applications and improving trial efficiency.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Neovascular age-related macular degeneration (nAMD) is a leading cause of vision loss.
- Accurate prediction of visual acuity is crucial for managing nAMD.
- Optical coherence tomography (OCT) provides detailed retinal imaging.
Purpose of the Study:
- To develop deep learning (DL) models for predicting best-corrected visual acuity (BCVA) from OCT images in nAMD patients.
- To assess the performance of DL regression and classification models.
Main Methods:
- Retrospective analysis of OCT images and BCVA from the HARBOR trial.
- Development of DL regression models to predict BCVA at concurrent and 12-month visits.
- Development of DL binary classification models to predict predefined BCVA thresholds.
Main Results:
- Regression models achieved R² values up to 0.84 for concurrent BCVA prediction.
- Classification models demonstrated high performance with AUC values up to 0.98.
- Models showed varying predictive accuracy for 12-month BCVA.
Conclusions:
- Deep learning models show significant potential for predicting BCVA from OCT in nAMD.
- Further research is needed to validate these models in clinical settings.
