Related Experiment Video
Updated: Aug 12, 2026
09:17
Using Retinal Imaging to Study Dementia
Published on: November 6, 2017
21.6K
Deep Learning to Discriminate Arteritic From Nonarteritic Ischemic Optic Neuropathy on Color Images
Ayse Gungor1,2, Raymond P Najjar3,4,5, Steffen Hamann6,7
1Sorbonne University, Paris, France.
JAMA Ophthalmology
|October 17, 2024
Summary
A deep learning system accurately distinguishes arteritic anterior ischemic optic neuropathy (AAION) from nonarteritic anterior ischemic optic neuropathy (NAION) using fundus images. This AI tool aids in preventing vision loss by improving diagnostic accuracy for AAION.
Area of Science:
- Ophthalmology
- Artificial Intelligence
- Medical Imaging
Background:
- Arteritic anterior ischemic optic neuropathy (AAION) diagnosis is crucial for preventing vision loss.
- Distinguishing AAION from nonarteritic anterior ischemic optic neuropathy (NAION) is challenging, especially without systemic symptoms or reliable biomarkers.
Purpose of the Study:
- To develop and validate a deep learning system (DLS) for differentiating AAION from NAION using color fundus images.
- To assess the DLS performance in acute-phase AAION detection.
Main Methods:
- An international study trained a DLS on 961 color fundus images from 802 patients with confirmed AAION or NAION.
- External validation was performed on a separate cohort using images from diverse fundus camera models.
Main Results:
- The DLS achieved high diagnostic performance on external testing: AUC of 0.97, sensitivity of 91.1%, specificity of 93.4%, and accuracy of 92.6%.
- The DLS significantly outperformed human expert accuracy (74.3% and 81.6%).
Conclusions:
- The DLS demonstrates excellent accuracy in discriminating AAION from NAION using only color fundus images.
- This AI tool has the potential to enhance clinical decision-making and reduce misdiagnosis, thereby improving patient outcomes.
