| Title |
Intraocular Corneal Ulcer Area Segmentation using Selective Image Processing |
| Authors |
김인환(Inhwan Kim) ; 김대원(Daewon Kim) |
| DOI |
https://doi.org/10.5573/ieie.2023.60.5.47 |
| Keywords |
Corneal ulcer; Flood-fill; Detection; Otsu's threshold; K-means clustering; Segmentation |
| Abstract |
Corneal ulcers are the most common symptoms of corneal diseases and require a professional examination. In addition, the subjective opinion of the medical staff may be included in determining whether an early onset or not. In this paper, the background and the corneal ulcer area were separated using various image preprocessing methods for the Intraocular image, and the light reflection area was treated with a labeling method. For each result of applying Otsu's threshold and K-means clustering algorithm, the corneal ulcer area was detected using the Flood-Fill algorithm, and both methods were selectively applied to derive a better image as the final result. As a result of comparing the application images of the two algorithms shown in the experimental results, the corneal ulcer region extracted from the Otsu's threshold algorithm application image was wider than the corneal ulcer region extracted by applying the K-means clustering algorithm, so this was used. In this study, we proposed a method for calculating the proportion of non-overlapping corneal ulcer regions in the corneal ulcer region extracted after application of Otsu's threshold algorithm in the two resulting images, and finally selecting an algorithm to be applied. As a result of the whole process, 90.54% accuracy was shown when measuring the Dice Coefficient, and the Mean Accuracy value showed an accuracy of 97.45%. In the future, we plan to continue our research to enable more accurate image segmentation using deep learning techniques. |