Lightweight Optic Disc and Optic Cup Segmentation Based on MobileNetv3 Convolutional Neural Network.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Glaucoma is a leading cause of blindness worldwide.
• Accurate optic disc (OD) and optic cup (OC) segmentation is crucial for glaucoma screening via the cup-to-disc ratio (CDR).
• Current deep learning methods often suffer from high computational costs and slow inference.

Purpose of the Study:

• To propose an efficient end-to-end method for OD and OC segmentation.
• To leverage a lightweight network (MobileNetv3) for improved computational efficiency.
• To enable precise and rapid glaucoma screening through accurate OD and OC segmentation.

Main Methods:

• Utilized MobileNetv3 as the core feature extractor for an end-to-end segmentation network.
• Incorporated boundary branches and adversarial learning for multi-label segmentation of OD and OC.
• Validated the method on three public datasets: Drishti-GS, RIM-ONE-r3, and REFUGE.

Main Results:

• Achieved high Dice coefficients for OD segmentation (0.974, 0.966, 0.962) and OC segmentation (0.900, 0.875, 0.880) across datasets.
• Demonstrated significant reductions in computational complexity and inference time compared to existing methods.
• Confirmed the method's effectiveness for efficient and precise OD and OC segmentation.

Conclusions:

• The proposed MobileNetv3-based approach offers an efficient and accurate solution for OD and OC segmentation.
• This method has the potential to enhance the speed and accessibility of glaucoma screening.
• The combination of lightweight architecture and adversarial learning provides a promising direction for medical image segmentation.