Particle Swarm Optimization-Based Approach for Optic Disc Segmentation.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Science

Background:

• Accurate optic disc segmentation is crucial for diagnosing ophthalmic diseases like glaucoma.
• Existing public fundus datasets suffer from limited image availability and uneven distribution.
• Current segmentation methods may require extensive data or pre-training, which is often infeasible.

Purpose of the Study:

• To propose a modified particle swarm optimization (PSO) algorithm for accurate optic disc segmentation.
• To address the challenges posed by limited and unevenly distributed fundus image datasets.
• To develop a data-efficient and pre-training-free segmentation method.

Main Methods:

• The optic disc segmentation problem is reformulated as an extreme value problem.
• Data preprocessing techniques are applied to enhance fundus image features and reduce noise.
• A modified PSO algorithm divides the search space into subgroups, enabling collaborative particle exploration.
• Particle fitness and contour accuracy are calculated using gradient values, with inter-subgroup particle attraction enhancing convergence.

Main Results:

• The proposed method demonstrated superior optic disc segmentation performance on the Drishti-GS and RIM-ONE V3 datasets.
• Substantial improvements in segmentation accuracy were observed compared to several state-of-the-art methods.
• The algorithm's effectiveness was validated despite the limitations of public fundus datasets.

Conclusions:

• The modified PSO algorithm offers a robust and accurate solution for optic disc segmentation.
• This approach effectively overcomes the challenges associated with small and unevenly distributed datasets.
• The proposed method holds significant potential for improving automated glaucoma diagnosis and other ophthalmic disease detection.