Related Experiment Video
Updated: Jun 29, 2025
08:57
Quantification of Optic Nerve Cross Sectional Area on MRI: A Novel Protocol using Fiji Software
Published on: September 4, 2021
4.0K
Recognition of Glaucomatous Fundus Images Using Machine Learning Methods Based on Optic Nerve Head Topographic
Chao-Wei Wu1, Tzu-Yu Huang2, Yeong-Cheng Liou2
1Department of Ophthalmology, Kaohsiung Medical University Hospital.
Journal of Glaucoma
|March 28, 2024
Summary
Machine learning effectively detects glaucoma from optic disc images, outperforming traditional rules. Automating feature extraction will streamline this diagnostic approach.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Glaucoma detection relies on optic disc topography.
- Clinical discriminant rules are currently used for diagnosis.
- Need for more effective and automated diagnostic tools.
Purpose of the Study:
- Compare machine learning classifiers with clinical discriminant rules.
- Evaluate the effectiveness of optic disc topographic features in glaucoma detection.
- Identify the most important topographic features for diagnosing glaucoma.
Main Methods:
- Retrospective case-control study of 800 fundus images (400 glaucoma, 400 non-glaucoma).
- Features extracted: vertical cup-to-disc ratio, optic rim width in four quadrants.
- Machine learning classifiers developed and compared against clinical rules.
Main Results:
- Machine learning classifiers significantly outperformed clinical discriminant rules.
- Extreme gradient boosting showed the highest performance in identifying glaucomatous images.
- Cup-to-disc ratio was the most critical feature; temporal optic rim width was least important.
Conclusions:
- Machine learning classifiers are effective for detecting glaucoma using optic disc topography.
- Automating feature extraction can create a straightforward and efficient diagnostic method.
- This approach offers a promising tool for glaucoma screening and diagnosis.
