Optic Disc Segmentation from Retinal Fundus Images via Deep Object Detection Networks.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Accurate optic disc segmentation is crucial for diagnosing ocular diseases.
• Current methods face challenges in precise segmentation of the optic disc.

Purpose of the Study:

• To propose a novel pipeline for optic disc segmentation from retinal fundus images.
• To leverage deep object detection networks for improved segmentation accuracy.

Main Methods:

• The optic disc segmentation problem is reframed as an object detection task.
• A Faster R-CNN object detector is employed to identify the optic disc.
• The predicted bounding box is transformed into a vertical, non-rotated ellipse to determine the optic disc boundary.

Main Results:

• The proposed method achieves state-of-the-art optic disc segmentation results.
• Performance was validated on the ORIGA dataset, outperforming existing techniques.
• The pipeline demonstrates high accuracy in segmenting the optic disc from retinal images.

Conclusions:

• The deep object detection approach offers a more straightforward and effective method for optic disc segmentation.
• This pipeline shows significant potential for enhancing computer-aided diagnosis of ocular diseases.
• The transformation of bounding boxes to ellipses provides accurate optic disc boundary determination.