Joint Optic Disc and Cup Segmentation Based on Multi-label Deep Network and Polar Transformation

Huazhu Fu, Jun Cheng, Yanwu Xu, Damon Wing Kee Wong, Jiang Liu, Xiaochun Cao

Abstract

Glaucoma is a chronic eye disease that leads to irreversible vision loss. The cup to disc ratio (CDR) plays an important role in the screening and diagnosis of glaucoma. Thus, the accurate and automatic segmentation of optic disc (OD) and optic cup (OC) from fundus images is a fundamental task. Most existing methods segment them separately, and rely on hand-crafted visual feature from fundus images. In this paper, we propose a deep learning architecture, named M-Net, which solves the OD and OC segmentation jointly in a one-stage multi-label system. The proposed M-Net mainly consists of multi-scale input layer, U-shape convolutional network, side-output layer, and multi-label loss function. The multi-scale input layer constructs an image pyramid to achieve multiple level receptive field sizes. The U-shape convolutional network is employed as the main body network structure to learn the rich hierarchical representation, while the side-output layer acts as an early classifier that produces a companion local prediction map for different scale layers. Finally, a multi-label loss function is proposed to generate the final segmentation map. For improving the segmentation performance further, we also introduce the polar transformation, which provides the representation of the original image in the polar coordinate system. The experiments show that our M-Net system achieves state-of-the-art OD and OC segmentation result on ORIGA dataset. Simultaneously, the proposed method also obtains the satisfactory glaucoma screening performances with calculated CDR value on both ORIGA and SCES datasets.

Introduction

Glaucoma is a leading cause of irreversible blindness, and early detection is vital for preserving vision. Clinicians primarily screen for the disease by assessing the Cup-to-Disc Ratio (CDR)—the ratio of the vertical diameter of the optic cup to the optic disc. Automating this assessment is challenging due to low contrast at the cup boundary, interference from pathological features like peripapillary atrophy (PPA), and the inherent spatial relationship where the cup is always nested within the disc.

Comparison of fundus images in normal and glaucomatous eyes, highlighting the optic disc and cup boundaries used for CDR calculation.

Most existing methods segment the disc and cup separately, ignoring their spatial dependencies. This paper introduces M-Net, a deep learning architecture designed for joint segmentation in a single stage, combined with a geometric preprocessing step called Polar Transformation to improve accuracy and handle class imbalance.

Methodology

The M-Net Architecture

M-Net is an end-to-end multi-label deep network that builds upon the U-Net architecture. Its key components include:

  • Multi-scale Input Layer: Feeds an image pyramid into the encoder, allowing the network to learn features at multiple receptive field sizes simultaneously.
  • U-shape Backbone: An encoder-decoder structure with skip connections to preserve high-resolution spatial information.
  • Side-output Layers: Act as "early classifiers" at different stages, providing deep supervision to help the model converge effectively.
  • Multi-label Dice Loss: Instead of multi-class classification, the task is treated as multi-label. Since the cup is inside the disc, a "cup" pixel is also a "disc" pixel. A Dice coefficient-based loss is used to handle the heavy class imbalance between the small cup and the background.
The overall pipeline and detailed architecture of the proposed M-Net system.

Polar Transformation

Before processing, the region around the optic disc is transformed from Cartesian to polar coordinates. This transformation converts the circular relationship between the cup and disc into a layered, horizontal relationship, making it easier for convolutional kernels to learn the "cup-inside-disc" constraint. Furthermore, it effectively "stretches" the central cup region, increasing its pixel representation from approximately 4% in Cartesian crops to 23% in polar crops, significantly reducing class imbalance.

Results and Discussion

The model was evaluated on the ORIGA dataset (650 images) and the SCES dataset (1,676 images). On ORIGA, M-Net with Polar Transformation achieved an overlapping error of 0.071 for the disc and 0.230 for the cup, outperforming traditional methods like superpixel classification.

For glaucoma screening, the system achieved an Area Under the Curve (AUC) of 0.8508 on ORIGA and 0.8998 on SCES based on the calculated CDR. The system is highly efficient, taking only 0.5 seconds to process an image on a GPU, making it suitable for large-scale population screening.

Conclusion

M-Net provides a robust, automated solution for joint optic disc and cup segmentation. By leveraging multi-scale features and geometric transformations, it achieves state-of-the-art results in CDR calculation and glaucoma screening. Future work may involve exploring more powerful backbone networks to handle extremely low-contrast images where boundaries are nearly invisible.

Summary updated

Abstract

Glaucoma is a chronic eye disease that leads to irreversible vision loss. The cup to disc ratio (CDR) plays an important role in the screening and diagnosis of glaucoma. Thus, the accurate and automatic segmentation of optic disc (OD) and optic cup (OC) from fundus images is a fundamental task. Most existing methods segment them separately, and rely on hand-crafted visual feature from fundus images. In this paper, we propose a deep learning architecture, named M-Net, which solves the OD and OC segmentation jointly in a one-stage multi-label system. The proposed M-Net mainly consists of multi-scale input layer, U-shape convolutional network, side-output layer, and multi-label loss function. The multi-scale input layer constructs an image pyramid to achieve multiple level receptive field sizes. The U-shape convolutional network is employed as the main body network structure to learn the rich hierarchical representation, while the side-output layer acts as an early classifier that produces a companion local prediction map for different scale layers. Finally, a multi-label loss function is proposed to generate the final segmentation map. For improving the segmentation performance further, we also introduce the polar transformation, which provides the representation of the original image in the polar coordinate system. The experiments show that our M-Net system achieves state-of-the-art OD and OC segmentation result on ORIGA dataset. Simultaneously, the proposed method also obtains the satisfactory glaucoma screening performances with calculated CDR value on both ORIGA and SCES datasets.
BibTeX
@article{fu2018joint,
  author = {Huazhu Fu and Jun Cheng and Yanwu Xu and Damon Wing Kee Wong and Jiang Liu and Xiaochun Cao},
  title = {{Joint Optic Disc and Cup Segmentation Based on Multi-label Deep Network and Polar Transformation}},
  journal = {IEEE Trans. Medical Imaging},
  year = {2018},
  url = {https://openreview.net/forum?id=6DBtgVyAded},
}
Figure