Assessment of Correlation Between Diabetic Retinopathy and Metabolic Biomarkers Using Artificial Intelligence

Area of Science:

• Endocrinology
• Ophthalmology
• Artificial Intelligence

Background:

• Diabetic retinopathy (DR) is a leading cause of blindness in diabetic patients.
• Early detection and treatment of DR are crucial for preventing vision loss.
• Artificial intelligence (AI) combined with nonmydriatic fundus cameras shows promise for DR screening.

Purpose of the Study:

• To systematically compare the diagnostic effectiveness of various nonmydriatic cameras and AI software for DR screening.
• To provide information for enhancing diabetes care plans and preventative actions in endocrine health.
• To address the knowledge gap in comparing nonmydriatic cameras and AI for DR detection.

Main Methods:

• A clinical study involving 900 diabetic patients with undiagnosed DR.
• Fundus images captured using three nonmydriatic cameras and assessed by EyeCheckup AI software.
• Comparison of AI diagnoses with evaluations by three retina specialists based on American Academy of Ophthalmology guidelines.

Main Results:

• Significant association found between DR prevalence and diabetes duration, HbA1c, CRP, and urinary albumin levels (p<0.05).
• AI demonstrated satisfactory performance in DR detection from fundus images.
• Sensitivity and specificity varied by camera: Canon CR2 AF (95.65%/95.92%), Topcon TRC-NW400 (95.19%/96.46%), Optomed Aurora (90.48%/97.21%).

Conclusions:

• Elevated HbA1c, CRP, urinary albumin, and longer diabetes duration are significant predictors of DR.
• Biomarkers like HbA1c, CRP, and urinary albumin can aid in proactive DR screening.
• Nonmydriatic fundus cameras with AI software offer potential for early DR detection and intervention.