Automatic Corneal Ulcer Segmentation Combining Gaussian Mixture Modeling and Otsu Method

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Corneal ulcers are a significant cause of vision impairment.
• Accurate segmentation of ulcer area is crucial for diagnosis and treatment monitoring.
• Manual segmentation is time-consuming and subjective.

Purpose of the Study:

• To develop and validate an accurate, automated pipeline for segmenting flaky corneal ulcer areas.
• To improve the robustness and accuracy of segmentation using the HSV color model.
• To compare the performance of the proposed pipeline with traditional RGB processing.

Main Methods:

• A joint segmentation approach combining Otsu's method and Gaussian Mixture Modeling (GMM).
• Intelligent determination of Gaussian components using information theory.
• Image processing in the HSV color space and grayscale with Gamma correction.
• Post-processing using morphological operations, median filtering, and hole-filling algorithms.

Main Results:

• The pipeline achieved a mean Dice Similarity Coefficient (DSC) of 0.88 on 150 clinical images.
• Processing in the HSV color space yielded significantly better results (mean DSC 0.88) than RGB (mean DSC 0.83).
• The automated segmentation closely matched the manually-delineated gold standard.

Conclusions:

• The proposed automated pipeline provides accurate and robust segmentation of corneal ulcers.
• Utilizing the HSV color model enhances segmentation performance compared to RGB.
• This method offers a promising tool for clinical applications in ophthalmology.