AI-Assisted Microfluidic Paper-Based Analytical Device with Au–Pt Nanoparticles for Multiplex, Interference-Resistant Quantification of Urinary Biomarkers

Urinary glucose, creatinine, and uric acid are vital biomarkers for diabetes and kidney disease management. However, multiplex point-of-care detection faces challenges due to insufficient sensitivity in complex urine matrices and signal cross-talk from interfering species. To address this, we developed an artificial intelligence (AI)-assisted microfluidic paper-based analytical device (μPAD) for simultaneous electrochemical quantification of these biomarkers. Screen-printed electrodes (SPEs) were modified with gold (Au)-platinum (Pt) bimetallic nanoparticles (NPs), leveraging their synergistic electrocatalysis to enhance hydrogen peroxide oxidation sensitivity by 25.4-fold. The optimized μPAD achieved rapid (<180 s) detection with low limits of detection (glucose: 10.1 μM; uric acid: 0.39 μM; creatinine: 141.7 μM) across clinically relevant ranges. Crucially, three multilayer perceptron (MLP) neural networks were applied to correct interference-induced errors, reducing mean absolute quantification errors from 30.9% to 6.3% (glucose) and 35.8% to 5.1% (creatinine). This integration of Au-Pt catalysis, pump-free μPAD design, and AI calibration enables highly sensitive and selective multiplex detection in urine. The developed platform demonstrates significant potential for decentralized diagnostics and longitudinal monitoring of diabetes and kidney injury.