1：Experimental Design and Method Selection:

The system involves preprocessing OCTA images for contrast enhancement and noise elimination using regional dynamic histogram equalization (RDHE) and a combination of generalized Gauss-Markov random field (GGMRF) model with adaptive gray level threshold estimation. Segmentation of retinal blood vessels from superficial and deep plexuses is performed using a technique combining prior intensity, current intensity, and higher-order spatial models, followed by Na?ve Bayes classification. Feature extraction includes calculating blood vessel density using Parzen window technique, vessel caliber, distance map of FAZ using region growing and morphological filters, and bifurcation points from vessel skeletons. A two-stage random forest classifier is used for diagnosis.

2：Sample Selection and Data Sources:

A dataset of 133 OCTA scans (34 normal and 99 mild DR cases) captured using a ZEISS AngioPlex OCT Angiography machine, with images of size 1024x1024 and 6x6 mm2 sections centered on the fovea.

3：List of Experimental Equipment and Materials:

ZEISS AngioPlex OCT Angiography machine for image acquisition; software for image processing and analysis.

4：Experimental Procedures and Operational Workflow:

Preprocess images, segment vessels, extract features (density, caliber, FAZ width, bifurcation points), and classify using RF classifier with k-fold cross-validation.

5：Data Analysis Methods:

Performance evaluated using accuracy, sensitivity, specificity, area under the curve (AUC), and Dice Similarity Coefficient (DSC). Cross-validation techniques (two-fold and four-fold) and comparison with other classifiers (SVM, KNN, classification tree).