Related Experiment Video
Updated: Apr 11, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.7K
The Application of a Deep Learning Algorithm for the Segmentation of Retinal Nerve Fiber Layer Across Different Optic
Roya Arian1, Milad Behzadi Far2, Reza Sadeghi2
1Department of Engineering, Durham University, Durham, UK.
Translational Vision Science & Technology
|April 10, 2026
Summary
A deep learning algorithm, RNFL-Net, accurately segments retinal nerve fiber layer (RNFL) thickness from optical coherence tomography (OCT) scans. This advanced method outperforms standard OCT devices, particularly in cases of optic disc edema.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Accurate measurement of retinal nerve fiber layer (RNFL) thickness is crucial for diagnosing and monitoring optic neuropathies.
- Standard optical coherence tomography (OCT) devices may have limitations in precisely segmenting RNFL, especially in conditions like optic disc edema.
- Deep learning (DL) offers potential for improving automated image analysis in ophthalmology.
Purpose of the Study:
- To evaluate the capability of a deep learning algorithm, RNFL-Net, to segment the RNFL from OCT scans.
- To assess RNFL-Net's performance in eyes with glaucomatous optic neuropathies and anterior optic neuropathies with optic disc edema.
- To compare the accuracy of RNFL thickness measurements obtained by RNFL-Net against manual segmentation and standard OCT devices.
Main Methods:
- RNFL-Net was developed with preprocessing steps including automatic blood vessel removal from peripapillary OCT B-scans.
- The algorithm was trained and validated on 1065 RNFL OCT B-scans and tested on 265 scans, with additional external validation datasets.
- The study included eyes from healthy controls, glaucoma patients, and patients with optic disc edema.
Main Results:
- RNFL-Net achieved high segmentation accuracy with a Dice coefficient of 0.95 on the validation set and 0.92 on the test set.
- In glaucoma cases, RNFL-Net demonstrated a lower Mean Absolute Error (MAE) of 6.21 µm compared to standard OCT (11.05 µm).
- For optic disc edema, RNFL-Net showed a significantly lower MAE of 13.04 µm versus standard OCT (22.94 µm), indicating superior performance.
Conclusions:
- RNFL-Net provides accurate RNFL segmentation in OCT scans.
- The DL algorithm exhibits superior performance in measuring RNFL thickness compared to standard OCT devices, especially in cases of optic disc edema.
- RNFL thickness measurements from RNFL-Net align well with ground truth data in both glaucoma and optic disc edema.
