| Title |
Corneal Ulcer Segmentation using Faster R-CNN and Image Processing |
| Authors |
김인환(Inhwan Kim) ; 김대원(Daewon Kim) |
| DOI |
https://doi.org/10.5573/ieie.2025.62.12.75 |
| Keywords |
Corneal Ulcer; Object detection; Faster R-CNN; Segmentation; k-means clustering; Active contour |
| Abstract |
Corneal ulcer is one of the common corneal diseases caused by infections with parasites such as bacteria, fungi, viruses, or acanthamoeba. It causes congestion, pain, and foreign body sensation of eyes, increasing of light sensitivity or lacrimal secretion, which can lead to blindness in severe cases. An examination by a medical staff is required to determine whether or not the disease has occurred, and in this case, an assistive role in the medical staff's judgment is necessary. In this paper, in order to make the distinction between the corneal ulcer area and other areas clear, the corneal ulcer area was marked as bounding box using Faster R-CNN, a RPN (Region Proposal Network) based object detection model. Subsequently we proposed a method of dividing the ulcer area through image processing techniques such as Active Contour and Labeling algorithm for images marked as box. The maximum Precision value of boxes extracted through Faster R-CNN was 94.80% and the Accuracy value for the corneal ulcer area segmented through the post-processing was 99.14%. |