Related Experiment Video
Updated: Jul 18, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
Prediction of Visual Field Progression with Baseline and Longitudinal Structural Measurements Using Deep Learning
Vahid Mohammadzadeh1, Sean Wu2, Sajad Besharati1
1From the Glaucoma Division, Stein Eye Institute, David Geffen School of Medicine, University of California Los Angeles (V.M., S.B., E.M., K.E., M.R., A.M., D.Z., J.C., K.N.-M.), Los Angeles, California, USA.
American Journal of Ophthalmology
|February 14, 2024
Summary
Deep learning accurately predicts glaucoma progression using retinal nerve fiber layer thickness and optic disc photographs. This tool can help identify patients at high risk for timely intervention.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Identifying glaucoma patients at high risk of progression is crucial for clinical management.
- Current methods often lack the predictive power needed for early intervention.
- Widely available structural data can potentially be leveraged for risk stratification.
Purpose of the Study:
- To develop and validate a deep learning (DL) algorithm for predicting visual field (VF) progression in glaucoma patients.
- To test the hypothesis that baseline or serial structural measures can predict VF progression.
Main Methods:
- A Siamese Neural Network with a ResNet-152 backbone was designed to predict VF progression.
- The model utilized serial optic disc photographs (ODP) and baseline retinal nerve fiber layer (RNFL) thickness from 3,079 eyes.
- Model performance was evaluated using the Area Under the ROC Curve (AUC) and tested on an external dataset.
Main Results:
- The DL model achieved an AUC of 0.813 incorporating baseline ODP and RNFL thickness.
- Adding serial ODPs improved prediction accuracy, with AUC reaching 0.894.
- The model demonstrated high accuracy (AUC=0.911) in predicting fast glaucoma progression and maintained performance on an external dataset (AUC=0.893).
Conclusions:
- Deep learning models can accurately predict visual field progression in glaucoma using readily available structural data.
- The developed DL algorithm, utilizing RNFL thickness and serial ODPs, shows potential as a clinical tool for risk stratification.
- Further validation is recommended for clinical implementation.
