Non-Invasive Precise Classification of Glomerular Diseases in Urine Based on Hyperspectral Technology

Glomerular diseases, characterized by primary glomerular injury, impose a significant global health burden. While renal biopsy remains the diagnostic gold standard, this study explores hyperspectral imaging (HSI) as a novel non-invasive methodology combining spectral and spatial analysis. Urine samples from patients with four glomerular disease subtypes (Minimal Change Disease, Diabetic Nephropathy, Membranous Nephropathy, IgA Nephropathy; 40 samples/subtype) underwent HSI acquisition. Using dimensionality-reduced HSI spectral data, we developed a ResNet-50 classification model. The model achieved high performance with 96.8% average five-fold cross-validation accuracy and a 0.982 AUC, confirming accurate multiclass differentiation feasibility from limited samples. Comparative analysis validated the superior efficacy of the integrated ResNet-50 and HSI approach for this classification task.