SCREEN-DR - Software Architecture for the Diabetic Retinopathy Screening.

Area of Science:

• Ophthalmology
• Medical Informatics
• Artificial Intelligence

Background:

• Diabetic retinopathy (DR) is a leading cause of preventable blindness.
• Early-stage DR often lacks symptoms, necessitating regular screening for timely intervention.
• Current DR screening programs can benefit from enhanced technological support and collaboration.

Purpose of the Study:

• To present the SCREEN-DR project's collaborative platform for a regional diabetic retinopathy screening program.
• To outline the roles of researchers and physicians in developing and utilizing the platform.
• To facilitate the creation of annotated datasets for artificial intelligence (AI) model training.

Main Methods:

• Researchers developed algorithms for image quality assessment, non-pathological case identification, lesion localization, and DR severity grading.
• Physicians annotated datasets, including precise visual delineation of lesions.
• The platform managed data collection, metadata indexing, dataset creation, and the annotation workflow.

Main Results:

• A functional collaborative platform was established, integrating research and clinical expertise.
• The platform supports advanced multimodal queries for annotated datasets.
• Export capabilities enable the feeding of AI algorithms with curated data for DR analysis.

Conclusions:

• The SCREEN-DR platform fosters a synergistic partnership between clinicians and researchers for effective DR screening.
• This collaborative approach enhances the development of AI tools for automated DR detection and grading.
• The platform facilitates the creation of high-quality, annotated datasets crucial for advancing diabetic retinopathy research and patient care.