A CNN-aided method to predict glaucoma progression using DARC (Detection of Apoptosing Retinal Cells)
Eduardo M Normando1,2, Tim E Yap1,2, John Maddison3
1ICORG, Imperial College London , London, UK.
Expert Review of Molecular Diagnostics
|April 21, 2020
Summary
An automated AI method accurately detects DARC spots to predict glaucoma progression. This AI-aided biomarker offers objective measurement for early detection and drug testing in glaucoma patients.
Area of Science:
- Ophthalmology
- Artificial Intelligence
- Biomarker Development
Background:
- Glaucoma poses a risk of rapid progression and blindness.
- Detection-of-Apoptosing-Retinal-Cells (DARC) is a novel method for visualizing apoptotic retinal cells.
- Accurate prediction of glaucoma progression is crucial for timely intervention.
Purpose of the Study:
- To develop an automated Convolutional Neural Network (CNN)-aided method for DARC spot detection.
- To utilize DARC spot detection for predicting glaucoma progression.
- To establish an objective biomarker for glaucoma management.
Main Methods:
- DARC images were obtained from healthy controls (n=40) and glaucoma patients (n=20).
- A CNN-aided algorithm was trained and validated using manual DARC spot counts from control subjects.
- The algorithm was tested on glaucoma eyes, comparing its performance to OCT-retinal nerve fiber layer measurements.
Main Results:
- The algorithm achieved 97.0% accuracy, 91.1% sensitivity, and 97.1% specificity in spot detection compared to manual grading in controls.
- In glaucoma patients, the algorithm showed 85.7% sensitivity, 91.7% specificity, and an AUC of 0.89.
- A significantly higher DARC count (p=0.0044) was observed in patients who later progressed.
Conclusions:
- The CNN-enabled algorithm provides an automated and objective measure of DARC.
- This AI-aided approach can serve as a biomarker for predicting glaucoma progression.
- The method facilitates the testing of new glaucoma drugs.
