Smartphone-integrated portable microfluidic platform for liver biomarker quantification using deep learning

Abstract Accurate and decentralized liver biomarker testing is critical for early diagnosis and monitoring of hepatic dysfunctions, particularly in resource-constrained settings. This work presents a novel smartphone-integrated colorimetric sensing platform that combines microfluidics, deep learning, and mobile health technologies to estimate liver biomarkers quantitatively. A stereolithography (SLA) 3D-printed microfluidic flow cell, optimized for low reagent use and high optical clarity, processes 100 µL of sample-reagent mixture via a peristaltic pump at 50 µL/s. Biomarker-specific chromogenic reactions are imaged within a controlled lighting enclosure using multiple smartphone models and analyzed using a convolutional neural network (CNN) for a regression approach. The system achieves clinically relevant detection ranges of 0.1–20 mg/dL for direct and total bilirubin, and 10–300 U/L for alanine aminotransferase (ALT) and aspartate aminotransferase (AST), with limits of detection of 0.1 mg/dL, 0.05 mg/dL, 2.97 U/L, and 2.5 U/L, respectively. A two-point smartphone adaptability framework ensures robust cross-device performance without retraining. An Android application has been developed, which provides users with disease identification, real-time inference, and visualization of result. This clinical-grade analyzer features an average coefficient of determination (R 2 ) of 0.997 for all biomarkers, and the repeatability is shown by coefficients of variation under 3%. This innovative, cost-effective and portable solution gives precise liver function assessment, making it ideal for rural healthcare and mobile diagnostics.