Automatic detection of the optic disc using majority voting in a collection of optic disc detectors

Abstract

This paper proposes an efficient method for locating the optic disc in retinal images automatically using majority voting scheme and data fusion. We show that instead of inventing a new algorithm which ends up being a minor variation on an old idea, the fusion of different optic disc (OD) detectors can enhance the overall performance of the detection system. The optic disc centre candidates of different optic disc detectors are marked in the image and a circular template is fit on each pixel in the image to count the outputs of these algorithms that fall within the radius. The location with maximum number of optic disc centre candidates is the hotspot and is used to localize the optic disc centre. An assessment of the performance of the combined optic disc detector versus detectors working separately is also presented. Our method achieved highest performance (overall 100% correct detection).

Generalizing the Majority Voting Scheme to Spatially Constrained Voting

“…The p n,k values calculated by (16) and the ones shown in Figure 6 are slightly differ. The reason of the difference is that in our geometric derivation to have a closed for, we have considered only disjoint discs that fall inside the ROI, as well.…”

“…For an impression of the problem, see Figure 1 showing the optic disc, and the region of interest (ROI) of the retinal image. Organizing more individual OD detector algorithms into a voting system may raise detection accuracy [16]. In our approach, all of the OD algorithms return with the OD center as a single pixel.…”

Detection of the optic disc in fundus images by combining probability models

“…In [16], we have proposed an ensemble-based single object detection system based on simple majority voting which outperforms the member detectors [10][11][12][13][14][15]. Here, the single outputs for the object center of the member algorithms are merged and the majority voting scheme is applied using a template of the shape of the object to detect its correct position.…”

“…To overcome the imperfectness of the individual algorithms, we study and adapt some of the state-of-the-art OD detectors and finally organize them into an ensemble framework in order to combine their strengths and maximize the accuracy of the localization of the OD. First in [16], we suggested an ensemble of them and tested a majority voting scheme with a circular template to detect the correct position of the OD center, where the individual algorithms had just a single candidate. As a further improvement of this model, in [17], we extracted more than one candidate for each algorithm to increase the chance of getting the OD location among them.…”

A Generalization of Majority Voting Scheme for Medical Image Detectors

“…From our empirical results we have found that if we choose different methods with complementary strengths for combining the detectors the overall performance of the system can substantially improve [2]. To be able to characterize the accuracy of this combined system a corresponding theoretical model is needed.…”