Methods and systems for optic nerve head segmentation

1. A method of classifying an optic nerve cup, a blood vessel, and optic nerve rim of an eye from a retinal image comprising:

receiving a retinal image;

determining a feature vector for a candidate pixel in the retinal image; and

classifying the candidate pixel as a cup pixel, a blood vessel pixel, or a rim pixel based on the feature vector using a trained classifier, wherein the classifier is at least one of: a k-Nearest Neighbors classifier; a linear discriminant classifier; and a quadratic discriminant classifier.

2. The method of claim 1 , wherein the retinal image is selected from the group consisting of:

a color retinal image;

a monochrome retinal image;

a retinal image reconstructed from a 3D volume set of retinal images;

a color stereo pair; and

a monochrome stereo pair.

3. The method of claim 1 , wherein training the classifier comprises:

presenting, to the classifier, a pre-classified pixel and a feature vector associated with the pre-classified pixel; and

determining, by the classifier, an optimal feature vector for the pre-classified pixel from a pre-determined set of features.

4. The method of claim 3 , wherein the pre-determined set of features comprises a non-stereo feature and a stereo disparity feature.

5. The method of claim 4 , wherein the non-stereo feature is selected from the group consisting of:

Gaussian derivatives of arbitrary scale;

order and orientation Gabor wavelets of arbitrary scale;

spatial frequency;

orientation;

texture feature based on intensity;

texture feature based on hue;

texture feature based on saturation;

texture feature based on variance;

color feature based on intensity;

color feature based on hue;

color feature based on saturation;

Hessian derivatives of arbitrary scale and eigenvalue sort order; and

difference of Gaussian convolution of different scales.

6. The method of claim 4 , wherein the stereo disparity feature is selected from the group consisting of:

depth map from stereo image;

physiologically plausible stereo shift estimation using Gabor wavelets; and

optic nerve head shape map from an imaging modality.

7. The method of claim 1 , further comprising:

identifying an optic nerve;

identifying an optic nerve cup and optic nerve rim; and

determining a cup-to-disc ratio.

8. A computer readable medium with computer readable code embodied thereon, for performing the steps of:

receiving a color retinal image;

determining a feature vector for a candidate pixel in the color retinal image; and

classifying the candidate pixel as a cup pixel, a blood vessel pixel, or a rim pixel based on the feature vector using a trained classifier, wherein the classifier is at least one of: a k-Nearest Neighbors classifier; a linear discriminant classifier; and a quadratic discriminant classifier.

9. The computer readable medium of claim 8 , wherein the color retinal image is selected from the group consisting of:

a color retinal image;

a monochrome retinal image;

a color stereo pair; and

a monochrome stereo pair.

10. The computer readable medium of claim 8 , wherein training the classifier comprises:

presenting, to the classifier, a pre-classified pixel and a feature vector associated with the pre-classified pixel; and

determining, by the classifier, an optimal feature vector for the pre-classified pixel from a pre-determined set of features.

11. The computer readable medium of claim 10 , wherein the pre-determined set of features comprises a non-stereo feature and a stereo disparity feature.

12. The computer readable medium of claim 11 , wherein the non-stereo feature is selected from the group consisting of:

Gaussian derivatives of arbitrary scale;

order and orientation Gabor wavelets of arbitrary scale;

spatial frequency;

orientation;

texture feature based on intensity;

texture feature based on hue;

texture feature based on saturation;

texture feature based on variance;

color feature based on intensity;

color feature based on hue;

color feature based on saturation;

Hessian derivatives of arbitrary scale and eigenvalue sort order; and

difference of Gaussian convolution of different scales.

13. The computer readable medium of claim 11 , wherein the stereo disparity feature is selected from the group consisting of:

depth map from stereo image;

physiologically plausible stereo shift estimation using Gabor wavelets; and

optic nerve head shape map from an imaging modality.

14. The computer readable medium of claim 8 , further comprising:

identifying an optic nerve;

identifying an optic nerve cup and optic nerve rim; and

determining a cup-to-disc ratio.

15. A system for classifying an optic nerve cup, a blood vessel, and an optic nerve rim of an eye from a color retinal image, comprising:

a computer readable medium; and

a processor programmed to:

receive a color retinal image,

determine a feature vector for a candidate pixel in the color retinal image, and

classify the candidate pixel as a cup pixel, a blood vessel pixel, or a rim pixel based on the feature vector using a trained classifier, wherein the classifier is at least one of: a k-Nearest Neighbors classifier; a linear discriminant classifier; and a quadratic discriminant classifier.

16. The system of claim 15 , wherein the color retinal image is selected from the group consisting of:

a color retinal image;

a monochrome retinal image;

a color stereo pair; and

a monochrome stereo pair.

17. The system of claim 15 , wherein training the classifier comprises:

presenting, to the classifier, a pre-classified pixel and a feature vector associated with the pre-classified pixel; and

determining, by the classifier, an optimal feature vector for the pre-classified pixel from a pre-determined set of features.

18. The system of claim 17 , wherein the pre-determined set of features comprises a non-stereo feature and a stereo disparity feature.

19. The system of claim 18 , wherein the non-stereo feature is selected from the group consisting of:

Gaussian derivatives of arbitrary scale;

order and orientation Gabor wavelets of arbitrary scale;

spatial frequency;

orientation;

texture feature based on intensity;

texture feature based on hue;

texture feature based on saturation;

texture feature based on variance;

color feature based on intensity;

color feature based on hue;

color feature based on saturation;

Hessian derivatives of arbitrary scale and eigenvalue sort order; and

difference of Gaussian convolution of different scales.

20. The system of claim 18 , wherein the stereo disparity feature is selected from the group consisting of:

depth map from stereo image;

physiologically plausible stereo shift estimation using Gabor wavelets; and

optic nerve head shape map from an imaging modality.

21. The system of claim 15 , wherein the processor is further programmed to perform the steps comprising:

identifying an optic nerve;

identifying an optic nerve cup and optic nerve rim; and

determining a cup-to-disc ratio.