Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

< Back to Article

Fig 1.

Introduction of OD and OC area.

(a) shows the structure of the optic nerve head. (b) shows the optic disc and cup structure.

More »

Fig 1 Expand

Fig 2.

The training and testing process.

More »

Fig 2 Expand

Fig 3.

The structure of RFC-Net.

Each blue block represents a recurrent block, each red block represents a 3 × 3 convolution, each orange block represents a 3 × 3 deconvolution.

More »

Fig 3 Expand

Fig 4.

Different variant of standard convolutional and recurrent blocks.

(a)Basic Units, (b)Recurrent Units, (c)Stack Recurrent Units, (d)Recurrent Basic Units, (e)Stack Recurrent Basic Units, (f)Unfold RCL Layers.

More »

Fig 4 Expand

Fig 5.

Polar transformation.

(a) and (b) represent retinal images in cartesian coordinates. (c) and (d) represent retinal images in polar coordinates. In (b) and (d), where yellow represents OC, red represents OD, and black represents background.

More »

Fig 5 Expand

Table 1.

Segmentation results with/without DA and PT.

More »

Table 1 Expand

Fig 6.

Visualization of the segmentation results before and after using polar transformation.

(a) Fundus retinal image. (b) The optic disc image extracted from (a). (c) The image after polar transformation. (d) Segmented image in polar coordinate system. (e) Segmented image in cartesian coordinate system. (f) Restore the segmented image under the optic disc image from (e). (g) Ground Truth. The green contour represent the boundary of OD, the blue contour represent the boundary of OC.

More »

Fig 6 Expand

Table 2.

Comparison of Pvalue analysis results based on F1 indicators.

More »

Table 2 Expand

Table 3.

Experiment results at F1 and BLE for the recurrent block in the Fig 4.

More »

Table 3 Expand

Table 4.

Experiment results at ACC, SEN and SPC for the recurrent block in the Fig 4.

More »

Table 4 Expand

Table 5.

Joint segmentation results at ACC, SEN and SPC for the recurrent block in the Fig 4.

More »

Table 5 Expand

Fig 7.

Comparison of segmented images with different recurrent units in the Drishti-GS1 dataset.

(a) Ground Truth. (b) Basic Units. (c) Recurrent Units. (d) Stack Recurrent Units. (e) Recurrent Basic Units. (f) Stack Recurrent Basic Units. The green contour represent the boundary of OD, the blue contour represent the boundary of OC.

More »

Fig 7 Expand

Fig 8.

Comparison of Receiver Operating Characteristic (ROC) curves of each structure in the Drishti-GS1 dataset.

(a) Cup. (b) Disc. (c) Optic.

More »

Fig 8 Expand

Table 6.

Experiment results of OD and OC segmentation on Drishti-GS1 dataset.

More »

Table 6 Expand

Fig 9.

This results demonstrates qualitative results of the proposed RFC-Net.

First column is the original image, second column is the ground truth, third column is the results of the RFC-Net(ours), four column is the results of the BCRF [18], and five column is the results of the Multiview [34]. The green contour represent the boundary of OD, the blue contour represent the boundary of OC.

More »

Fig 9 Expand

Table 7.

F1, BLE, accuracy, sensitivity and specificity with the RFC-Net, FCN, U-net, M-Net and CE-Net models on Drishti-GS1 dataset.

More »

Table 7 Expand

Fig 10.

Ground truth, our methods RFC-Net, FCN, U-Net, M-Net and CE-Net real segmentation contours.

(a) Ground Truth. (b)RFC-Net. (c)FCN. (d)U-Net. (e)M-Net. (f)CE-Net.

More »

Fig 10 Expand

Table 8.

Ablation study of each module on Drishti-GS1 dataset.

More »

Table 8 Expand