Optic Disk Detection in Fundus Image Based on Structured Learning.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Automated optic disk (OD) detection is crucial for computer-aided diagnosis of eye diseases.
• Accurate segmentation of the OD is essential for quantitative analysis and disease monitoring.

Purpose of the Study:

• To propose a novel algorithm for automated optic disk detection using structured learning.
• To evaluate the algorithm's performance on public datasets and compare it with state-of-the-art methods.

Main Methods:

• A classifier model is trained using structured learning.
• The trained model generates an edge map of the optic disk.
• Thresholding and Circle Hough Transform are applied for final OD segmentation.

Main Results:

• The algorithm achieved promising results on three public datasets.
• Key performance metrics include an area overlap coefficient of 0.8605 and Dice coefficient of 0.9181.
• High accuracy (0.9777) and favorable true positive (0.9183) and false positive (0.0102) fractions were reported.

Conclusions:

• The proposed structured learning-based algorithm is a competitive and reliable method for optic disk segmentation.
• This automated approach can significantly aid in the development of computer-aided systems for eye disease detection.