Algorithm 1: RETFound-Based DRIL Classification Architecture.
1:	Input: OCT images with foveal annotations
2:	Model: RETFound (Vision Transformer)
3:	// Phase 1: Preprocessing the OCT Images and extracting central foveal 1 mm region
4:	for all class $c\in \{\mathrm{DRIL},\mathrm{No}-\mathrm{DRIL}\}$ do
5:	for all image $I_i$ in class c do
6:	Load Ii and annotation file
7:	Read foveal center coordinates $f_x$
8:	$I_{crop}\leftarrow \mathrm{CropCentral}(I_i,f_x,\mathrm{width}=1\mathrm{mm})$
9:	$I_{enhanced}\leftarrow \mathrm{CLAHE}\left(I_{crop}\right)$
10:	Save $I_{enhanced}$
11:	end for
12:	end for
13:	// Phase 2: Splitting the dataset for training and testing
14:	$\mathcal{D}\leftarrow$ Read all preprocessed images
15:	Shuffle $\mathcal{D}$ with seed = 42
16:	${\mathcal{D}}_{train},{\mathcal{D}}_{test}\leftarrow \mathrm{Split}(\mathcal{D},0.85,0.15)$
17:	// Phase 3: Cross-validation and model training
18:	Initialize $K=5$
19:	for    $k=1$ to K do
20:	Split ${\mathcal{D}}_{train}$ into Train and Validation sets
21:	Initialize RETFound with pretrained weights
22:	Freeze backbone layers
23:	Train classification head
24:	Unfreeze layers for fine-tuning
25:	Validate on Validation set
26:	end for
27:	// Phase 4: Model selection and evaluation
28:	Select best model based on cross-validation performance
29:	Evaluate final fold-specific trained models on the held-out test set
30:	Compute-AC, SE, SP, AUC, MCC and Plot Visualizations and other curves