Algorithm 2. Unsupervised clustering of RBC latent features using k-means

RBC_Clustering_Encoder(samples, encoder_models, num_clusters_list)   for each sample_id in samples do     initialize paths for model, images, and outputs     create output folders if not exist     load pre-trained encoder model for current sample_id     load and preprocess RBC images   → x_test     normalize pixel values (0–1)     encode images using encoder     → encoded_imgs     remove existing clustering score CSV if exists     for each num_clusters in num_clusters_list do       apply KMeans clustering (num_clusters)       compute clustering labels     → labels       calculate Silhouette Score     → sil_score       calculate Davies–Bouldin Index  → dbi_score       save scores to CSV log       create cluster folders and copy images based on labels       plot silhouette visualization per cluster       plot metrics comparison (Silhouette vs. DBI)       save plots       apply UMAP to reduce encoded_imgs to 2D       plot and save UMAP scatter plot with cluster coloring       save plots     end for     compile per-sample clustering report summarizing metrics and plots     append results to global clustering summary (multi-sample CSV)   end for   compute total execution time and display summary   return clustering_results_summary