EyeHealer: A large-scale anterior eye segment dataset with eye structure and lesion annotations

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Anterior segment eye diseases are common, necessitating efficient clinical diagnosis.
• Automated tools for segmenting eye lesions can significantly improve clinical care efficiency.
• Medical image annotation datasets are scarce, hindering deep learning model development.

Purpose of the Study:

• To develop a novel medical image annotation system, EyeHealer.
• To create a large-scale anterior eye segment dataset with pixel-level annotations.
• To establish a system for automated classification and segmentation of anterior eye diseases.

Main Methods:

• Developed the EyeHealer annotation system.
• Created a large-scale dataset of anterior eye segments with annotated structures and lesions.
• Conducted experiments using deep learning models for disease classification and lesion segmentation.

Main Results:

• The EyeHealer dataset contains pixel-level annotations for eye structures and lesions.
• Semantic segmentation models demonstrated superior performance compared to medical segmentation models.
• The developed system shows promise for automated classification and segmentation tasks.

Conclusions:

• The EyeHealer system and dataset facilitate research in automated anterior eye disease analysis.
• Publicly releasing the dataset will encourage further advancements in the field.
• Deep learning models, particularly semantic segmentation, show potential for clinical applications in ophthalmology.