Artificial Intelligence for the Detection of Diabetic Retinopathy in Primary Care: Protocol for Algorithm Development

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Imaging

Background:

• Diabetic retinopathy (DR) is a leading cause of blindness globally.
• Early detection through regular eye examinations is crucial for managing DR.
• Convolutional neural networks show promise in identifying DR with high accuracy.

Purpose of the Study:

• Develop an artificial intelligence (AI) algorithm for DR detection in diabetic patients.
• Scientifically validate the AI algorithm for use as a primary care screening tool.
• Compare AI performance against current diagnostic methods.

Main Methods:

• Two concurrent studies: AI algorithm development and a prospective diagnostic capacity study.
• AI model training using 80% of images, with 20% for validation.
• Comparison of AI performance against a double-blind reading by retina specialists.

Main Results:

• Image data collection commenced in late 2018.
• AI algorithm development expected to take 3-4 months.
• Patient cohort inclusion and preliminary results anticipated by mid-2019.

Conclusions:

• The developed AI algorithm is expected to match or surpass current DR screening performance.
• The AI tool can serve as a valuable complement or alternative to existing screening methods.
• This advancement aims to improve DR management and reduce blindness.