| Title |
Corneal Ulcers Detection Using Random Seed Appointment Algorithm |
| Authors |
임진혁(Jinhyuk Im) ; 김대원(Daewon Kim) |
| DOI |
https://doi.org/10.5573/ieie.2019.56.9.53 |
| Keywords |
Medical image processing, Corneal ulcer detection, Flood-Fill, Otsu’s algorithm. |
| Abstract |
In modern medicine image processing has become indispensable in the treatment and diagnosis of patients. In particular, image segmentation algorithms are used in medical imaging to detect diseased areas and measure their size There are many image segmentation algorithms. One of the typical methods is to enlarge a range by applying a specific threshold value based on a reference point. However, since images involving corneal ulcers do not always contain a certain shape of pattern or color accuracy cannot be guaranteed when various types of ulcer images are distinguished when segmentation is carried out with algorithms using a fixed threshold value. This study involved the development of an effective detection algorithm for corneal ulcers. The approach can be applied to images of various ulcer types by setting a reference point automatically through the application of the Otsu’s algorithm, minimizing user interventions and making the classification of the ulcer area easy. Preprocessing such as gray scale transformations, medium filters, histogram equalization, and gamma correction of corneal ulcer images can be applied. To evaluate this detection algorithm, the average accuracy and Dice's coefficient were utilized as evaluation indices, and the results were compared with those achieved using an edgeless active contour method, which showed a superior performance among existing image segmentation algorithms.. The method proposed in this study showed an improved accuracy of up to 10% or higher compared with that of the comparison group and the flood-till and Otsu’s algorithms. The performance is expected to improve through various follow-on studies in the future. |