Real-World Evaluation of Artificial Intelligence-Based Diabetic Retinopathy Screening Using the Optomed Aurora...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Diabetic retinopathy (DR) is a leading cause of preventable blindness.
• Early detection through screening is crucial for preventing vision loss in diabetic patients.
• Handheld fundus cameras offer potential for accessible DR screening.

Purpose of the Study:

• To evaluate the diagnostic performance of the Aireen AI algorithm for grading diabetic retinopathy.
• To assess the accuracy of AI-assisted DR grading using images from the handheld Optomed Aurora fundus camera.

Main Methods:

• A total of 624 fundus images were graded for diabetic retinopathy by two retina specialists and the Aireen AI algorithm.
• Performance metrics including sensitivity, specificity, and diagnostic accuracy were calculated against expert ophthalmologist grading.
• Image sufficiency for DR classification was assessed.

Main Results:

• 97% of captured fundus images were deemed sufficient for DR classification.
• The Aireen AI algorithm achieved 94.8% sensitivity and 91.4% specificity for DR detection.
• Overall diagnostic accuracy for DR was reported at 92.7%.

Conclusions:

• The Aireen AI algorithm demonstrates high diagnostic accuracy for diabetic retinopathy detection in images from the Optomed Aurora handheld camera.
• AI combined with handheld fundus technology shows promise for streamlining DR screening, reducing healthcare professional burden, and improving patient outcomes.
• This validated approach may enhance accessibility and efficiency in diabetic retinopathy screening programs.