Deep level set method for optic disc and cup segmentation on fundus images.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Glaucoma is a primary cause of irreversible blindness globally.
• Accurate segmentation of the optic cup (OC) and optic disc (OD) is vital for glaucoma screening.
• Current screening methods often rely on the vertical cup-to-disc ratio and other clinical indicators.

Purpose of the Study:

• To develop and evaluate a deep level set method for precise segmentation of the optic cup and optic disc.
• To enhance the accuracy of glaucoma screening through improved automated segmentation.
• To assess the method's performance on benchmark datasets.

Main Methods:

• A multi-scale convolutional neural network was employed to predict initial contours and evolution parameters for level set segmentation.
• The network integrated augmented prior knowledge and was supervised using active contour loss for refined segmentation.
• The deep level set approach was validated on the REFUGE and Drishthi-GS1 datasets.

Main Results:

• The proposed method achieved high Intersection over Union (IoU) scores: 93.61% for the optic cup and 96.69% for the optic disc on the REFUGE dataset.
• Segmentation results on the Drishthi-GS1 dataset demonstrated superior performance compared to state-of-the-art methods.
• The method effectively refines initial contours using predicted evolution parameters for detailed boundary accuracy.

Conclusions:

• The deep level set method offers a robust and accurate approach for optic cup and optic disc segmentation.
• This technique shows significant potential for improving automated glaucoma screening and diagnosis.
• The proposed method outperforms existing state-of-the-art techniques in segmenting key ocular structures.