AUTOMATIC CUP-TO-DISC RATIO MEASUREMENT SYSTEM

• US 20110091083A1
• Filed: 05/14/2008
• Published: 04/21/2011
• Est. Priority Date: 05/14/2008
• Status: Active Grant

• Current Assignee: Agency For Science Technology and Research (Government of Singapore), Singapore Health Services Pte Limited (MOH Holdings Pte Limited), National University of Singapore
• Original Assignee: Agency For Science Technology and Research (Government of Singapore), Singapore Health Services Pte Limited (MOH Holdings Pte Limited), National University of Singapore
• Inventors: Li, Huiqi, Wong, Wing Kee, Liu, Jiang, Lim, Joo Hwee, Wong, Tien Yin
• Granted Patent: US 8,705,826 B2
• Time in Patent Office: Days
• US Class Current: 382/128
• CPC Class Codes: A61B 3/12 for looking at the eye fundus, e.g. ophthalmoscopes A61B3/13 takes precedence
• CPC Class Codes: G06T 2207/10101 Optical tomography; Optical coherence tomography [OCT]
• CPC Class Codes: G06T 2207/20084 Artificial neural networks [ANN]
• CPC Class Codes: G06T 2207/20221 Image fusion; Image merging
• CPC Class Codes: G06T 2207/30041 Eye; Retina; Ophthalmic
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/11 Region-based segmentation

Abstract

A two-dimensional retinal fundus image of the retinal fundus of an eye is processed by optic disc segmentation (2) followed by cup segmentation 4. Data derived from the optic disc segmentation (i.e. the output of the disc segmentation (2) and/or data derived from the output of the optic disc segmentation step, e.g. by a smoothing operation 3) and data derived from the out-put of the optic cup segmentation (i.e. the output of the cup segmentation (4) and/or data derived from the output of the optic disc segmentation, e.g. by a smoothing operation 5) are fed (6) to an adaptive model which has been trained to generate from such inputs a value indicative of cup-to-disc ratio (CDR) of the eye. The CDR is indicative of glaucoma. Thus, the method can be used to screen patients for glaucoma.

Claims

1. A method for obtaining a CDR value from a single 2D retinal funds image, comprising the steps of:

(i) performing optic disc segmentation;

(ii) performing optic cup segmentation; and

(iii) inputting data derived from the optic disc segmentation and data derived from the optic cup segmentation into an adaptive model operative to generate a CDR value.

2. A method according to claim 1 in which, following said step (i), the result of the optic disc segmentation is subject to a smoothing step to produce smoothed disc segmentation data which is employed in said step (ii), and as data input to said adaptive model.

3. A method according to claim 1 in which, following said step (ii), one or more smoothing algorithms are applied to data output from said optic cup segmentation step (ii), to generate respective sets of smoothed optic cup data, said set or sets of smoothed optic cup data being input to said adaptive model in step (iii).

4. A method according to claim 3 in which there are a plurality of said smoothing algorithms, generating a plurality of respective sets of smoothed optic cup data which are input to said adaptive mode in step (iii)

5. A method according to claim 3 in which both said smoothed optic cup data and said data output from said optic cup segmentation step (ii) are input to said adaptive model.

6. A method according to claim 1 in which said step (i) of optic disc segmentation includes a preliminary step of identifying a region-of-interest in said fundus image based on the intensity of points in the image.

7. A method according to claim 1 in which said step (i) of optic disc segmentation employs a variational level set algorithm.

8. A method according to claim 1 in which said step (ii) of optic cup segmentation includes a first sub-step of estimating the cup boundary using a threshold technique, followed by a second sub-step employing a level set technique.

9. A method according to claim 1 in which said step (ii) of optic cup segmentation is performed by a plurality of methods, each method generating a respective optic cup segmentation, step (iii) including inputting data derived from each of these optic cup segmentations into the adaptive model.

10. A method according to claim 1 in which the fundus image contains a plurality of colour components, and at least one of said segmentation steps (i) and (ii) is performed using only a single components of the colour image.

11. A method according to claim 1 in which said adaptive model is a multi-layer neural network.

12. A method according to claim 1 including a preliminary stage of training said adaptive model using a training set comprising target CDR values generated by a clinician.

24. A computer program product on which are stored program instructions readable by a computer to cause the computer to carry out the method according to claim 1.

13. An apparatus for obtaining a CDR value from a single 2D retinal fundus image, comprising a processor arranged to perform the steps of

(i) performing optic disc segmentation;

(ii) performing optic cup segmentation; and

(iii) inputting data derived from the optic disc segmentation and data derived from the optic cup segmentation into an adaptive model operative to generate a CDR value.

14. An apparatus according to claim 13 in which the processor is arranged, following said steps (i), to smooth the result of the optic disc segmentation to produce smoothed disc segmentation data which is employed as data input to said adaptive model.

15. An apparatus according to claim 13 in which the processor is arranged, following said step (ii), to perform one or more smoothing algorithms to data output from said optic cup segmentation step (ii), to generate respective sets of smoothed optic cup data, and to input said set or sets of smoothed optic cup data to said adaptive model in step (iii)

16. An apparatus according to claim 15 in which the processor is arranged to perform a plurality of said smoothing algorithms, generating a plurality of respective sets of smoothed optic cup data which are input to said adaptive model in step (iii).

17. An apparatus according to claim 15 in which both said smoothed optic cup data and said data output from said optic cup segmentation step (ii) are input to said adaptive model.

18. An apparatus according to claim 13 in which said processor is arranged, as part of said step (i) of optic disc segmentation, to perform a preliminary step of identifying a region-of-interest in said fundus image based on the intensity of points in the image.

19. An apparatus according to claim 13 in which said processor is arranged to employ a variation level set algorithm in step (i) of optic disc segmentation.

20. An apparatus according to claim 13 in which said processor is arranged, in step (ii), to perform a first sub-step of estimating the cup boundary using a threshold technique, followed by a second sub-step employing a level set technique.

21. An apparatus according to claim 13 in which said processor is arranged, to perform said step (ii) of optic cup segmentation by a plurality of methods, each method generating a respective optic cup segmentation, step (iii) including inputting data derived from each of these optic cup segmentations into the adaptive model.

22. An apparatus according to claim 13 in which the processor is arranged to receive said fundus image as a plurality of colour components, and to perform at least one of said segmentation steps (i) and (ii) using only a single component of the colour image.

23. an apparatus according to claim 13 in which said adaptive model is a multi-layer neural network.