Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2009 Dec;13(6):859-70.
doi: 10.1016/j.media.2009.08.003. Epub 2009 Sep 4.

Fast detection of the optic disc and fovea in color fundus photographs

Meindert Niemeijer  1 , Michael D Abràmoff, Bram van Ginneken
Affiliations

Fast detection of the optic disc and fovea in color fundus photographs

Meindert Niemeijer et al. Med Image Anal. 2009 Dec.

Abstract

A fully automated, fast method to detect the fovea and the optic disc in digital color photographs of the retina is presented. The method makes few assumptions about the location of both structures in the image. We define the problem of localizing structures in a retinal image as a regression problem. A kNN regressor is utilized to predict the distance in pixels in the image to the object of interest at any given location in the image based on a set of features measured at that location. The method combines cues measured directly in the image with cues derived from a segmentation of the retinal vasculature. A distance prediction is made for a limited number of image locations and the point with the lowest predicted distance to the optic disc is selected as the optic disc center. Based on this location the search area for the fovea is defined. The location with the lowest predicted distance to the fovea within the foveal search area is selected as the fovea location. The method is trained with 500 images for which the optic disc and fovea locations are known. An extensive evaluation was done on 500 images from a diabetic retinopathy screening program and 100 specially selected images containing gross abnormalities. The method found the optic disc in 99.4% and the fovea in 96.8% of regular screening images and for the images with abnormalities these numbers were 93.0% and 89.0% respectively.

PubMed Disclaimer

Figures

Fig. 1
Fig. 1
A digital color fundus photograph with a circle marking the optic disc (left) and another circle in the center of which is the fovea. The vascular arch marked in the image is formed by the major arteries and veins that leave the optic disc up- and downwards.
Fig. 2
Fig. 2
a) The template used to extract features for localization of the optic disc. b) The template used to extract feature for localization of the fovea.
Fig. 3
Fig. 3
The optic disc template placed in a fundus image and two derived images. The points where unique vessel segments cross the template border are indicated by the blue dots. a) The green plane of the color fundus photograph. b) The vessel probability map. c) The skeletonized vessel segmentation. The vessel centerline points that lie on the template border, marked by the dots, are the locations where the local vessel width and orientation based features are measured.
Fig. 4
Fig. 4
A graph plotting the average prediction error ε̄ in pixels against the k parameter value of the kNN regressor in part of the training set.
Fig. 5
Fig. 5
The optic disc detection step intermediate results. In the distance maps the border of the FOV is indicated by the dashed line. a) A color fundus photograph. b) The 10 × 10 optic disc search grid, sampling points are represented by the bright pixels c) Blurred distance map, result of rough grid search for the optic disc center (lower intensity equals lower distance). d) Blurred distance map of fine grid search for the optic disc center.
Fig. 6
Fig. 6
The fovea detection step intermediate results. In the distance maps the border of the FOV is indicated by the dashed line. a) The foveal search area (within the rectangle) based on the location of the optic disc, inside the search area each sampling position is indicated with a single dot. In this example the search area to the left of the optic disc lies outside the FOV and is therefore not processed or shown. b) After blurring, the pixel with the lowest value indicates the fovea location. c) The color fundus image from Figure 5a with the results superimposed. The circles are centered on the found locations.
Fig. 7
Fig. 7
A boxplot for both the proposed method (PM) and second observer (2O), of the cases in which ε was lower than 50 pixels. Indicated are the min, max, the median (white line), the mean (red diamond), the first and third quartile. To derive a meaningful boxplot from the data, outliers with a positioning error greater than 50 pixels were removed. From left to right 2, 2, 4, 0, 0, 0, 32, 15, 11, 7, 7 and 3 cases were removed.
Fig. 8
Fig. 8
Example results, in each image the locations detected by the proposed method are shown by the dot (optic disc) and the square (fovea). The two circular objects in the image, one around the optic disc and one around the fovea, represent the areas of the image that are used for the detection accuracy determination. That is, whenever the algorithm detects the appropriate location anywhere within the indicated regions it is counted as a correct detection. The circles around the optic disc indicate the border of the optic disc while the fovea circles have a standard size (radius is 40 pixels). Differences in the apparent size of the fovea circles between images are due to scaling for display. a-d) Representative, correct results from Set 2. e-f) Representative, correct results from Set 3. g) No markers are shown due to complete failure of the method. h) The optic disc was detected successfully but the algorithm failed to identify the fovea position. i) Again the optic disc was found successfully but the fovea detection failed.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Niemeijer M, van Ginneken B, Russel S, Suttorp-Schulten M, Abràmoff MD. Automated detection and differentiation of drusen, exudates, and cotton-wool spots in digital color fundus photographs for diabetic retinopathy diagnosis. Investigative Ophthalmology & Visual Science. 2007;48:2260–2267. - PMC - PubMed
    1. Abràmoff MD, Alward WLM, Greenlee EC, Shuba L, Kim CY, Fingert JH, Kwon YH. Automated segmentation of the optic disc from stereo color photographs using physiologically plausible features. Invest Ophthalmol Vis Sci. Apr, 2007. pp. 1665–1673. [Online]. Available: http://dx.org/10.1167/iovs.06-1081. - DOI - PMC - PubMed
    1. Early Treatment Diabetic Retinopathy Study Research Group. Early Photocoagulation for Diabetic Retinopathy: ETDRS report 9. Ophthalmology. 1991;98:766–785. - PubMed
    1. Klonoff D, Schwartz D. An economic analysis of interventions for diabetes. Diabetes Care. 2000;23(no 3):390–404. - PubMed
    1. Abràmoff MD, Niemeijer M, Suttorp-Schulten MSA, Viergever MA, Russell SR, van Ginneken B. Evaluation of a system for automatic detection of diabetic retinopathy from color fundus photographs in a large population of patients with diabetes. Diabetes Care. 2008;31(no 2):193–198. - PMC - PubMed

Publication types