Related Experiment Video
Updated: May 18, 2026
09:17
Using Retinal Imaging to Study Dementia
Published on: November 6, 2017
Deep learning strategies for estimating retinal thickness from fundus images: a comparative study with multi-device
Noriyoshi Takahashi1, Nikhil Gadiraju1, Judy E Kim1
1Department of Ophthalmology, University of Texas Southwestern Medical Center, Dallas, TX, USA.
Scientific Reports
|May 16, 2026
Summary
This study developed a deep learning model to estimate retinal thickness from color fundus photographs, eliminating the need for costly optical coherence tomography scans in diabetic retinopathy screening.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Color fundus photographs (CFPs) are accessible for diabetic retinopathy (DR) screening.
- Optical coherence tomography (OCT) offers detailed retinal structure but is less practical for large-scale screening due to cost and complexity.
Purpose of the Study:
- To develop a deep learning framework using a foundation model (FM) to estimate OCT-derived total retinal thickness (TRT) maps directly from CFPs.
- To evaluate different decoder training strategies for TRT estimation and assess robustness to inter-device variability.
Main Methods:
- Utilized the RETFound foundation model as the backbone.
- Trained and evaluated on the AI-READI dataset with paired CFPs and OCT scans from Topcon Maestro2 and Triton devices.
- Compared Individual, Combination, and Random Selection decoder training strategies for TRT estimation.
Main Results:
- A significant inter-device thickness difference was observed between Maestro2 and Triton OCT devices.
- Individual and Combination models showed low bias (approx. 5% relative difference) against device-specific references.
- The Random Selection model demonstrated robustness to device variability by converging to an average TRT distribution.
Conclusions:
- The proposed FM-based framework successfully estimates retinal thickness from CFPs, bridging the gap between CFP and OCT imaging.
- This approach offers a cost-effective and practical solution for large-scale DR screening by leveraging readily available CFPs.