Multiparametric OCTA Biomarkers for Classifying Mild Diabetic Retinopathy: A Cross-Sectional Evaluation

Area of Science:

• Ophthalmology
• Medical Imaging
• Biomedical Engineering

Background:

• Diabetic retinopathy (DR) is a leading cause of vision loss.
• Early detection of mild DR is crucial for timely intervention.
• Optical coherence tomography angiography (OCTA) provides detailed vascular imaging.

Purpose of the Study:

• To assess if combining OCTA features enhances discrimination between mild DR and healthy controls (HCs).
• To identify stable OCTA features that contribute to high-performing diagnostic models.

Main Methods:

• Quantitative OCTA parameters were extracted from superficial and deep vascular plexuses.
• Support vector machine (SVM) models were used to evaluate feature combinations.
• Model performance was assessed using repeated cross-validation, measuring the area under the ROC curve (AUC).

Main Results:

• Twenty-seven OCTA parameters showed significant differences between mild DR and HCs.
• Single-parameter models had moderate discriminative performance (AUC 0.47-0.88).
• Multiparameter combinations achieved higher discrimination, with some reaching AUCs around 0.95.

Conclusions:

• Multiparametric OCTA analysis improves the discrimination between mild DR and HCs.
• Feature stability and complementarity, not just individual strength, are vital for robust diagnostic models.