Implementation of Artificial Intelligence-Based Diabetic Retinopathy Screening in a Tertiary Care Hospital in Quebec:...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Diabetic retinopathy (DR) affects approximately 25% of individuals with diabetes in Canada.
• Early detection of DR is critical to prevent vision loss.

Purpose of the Study:

• To evaluate the real-world performance of an artificial intelligence (AI) system for diabetic retinopathy (DR) screening.
• Assess the diagnostic accuracy and cost-effectiveness of the CARA AI system in a clinical setting.

Main Methods:

• Prospective recruitment of adult patients with diabetes at a tertiary care center.
• Dual-pathway screening using the Computer Assisted Retinal Analysis (CARA) AI system and standard ophthalmological examination.
• Measurement of AI system's sensitivity and specificity for referable DR, any retinopathy, and diabetic macular edema (DME).

Main Results:

• CARA demonstrated 87.5% sensitivity and 66.2% specificity for detecting referable DR at the patient level.
• At the eye level, CARA achieved 88.2% sensitivity and 71.4% specificity for any retinopathy.
• For diabetic macular edema (DME) detection, CARA showed 100% sensitivity and 81.9% specificity. Potential annual savings of CAD $245,635 were estimated.

Conclusions:

• Semiautomated AI system integration for DR screening shows high sensitivity in real-world application.
• The CARA system has the potential to enhance screening efficiency and reduce healthcare costs.
• Further validation studies are recommended to confirm the AI system's utility.