Microfluidic pH monitoring of trace urine samples via optimized optical layout and machine learning

Microfluidic chip monitors 1.9 μL urine pH from 4.0 to 9.0. • Fiber-collimated optics and ML resist urine pigment, turbidity, hemolysis. • GBDT-LRGB model is optimized and verified with over 2000 groups of data. • 92.5% samples show pH residuals < 0.2, within clinical 0.5 deviation. • Adaptable for organ-on-a-chip, single-cell culture, and IVD devices. The demand for stable and accurate continuous pH monitoring of trace liquids in microfluidic chips keeps rising, yet existing methods fail to balance five core indicators: stability, accuracy, measurement range, anti-interference and low cost. This work constructs a microfluidic pH detection system via optimized optical layout, 3D microfluidic structure and machine learning. Fiber coupling and collimating lenses homogenize light inside the micro-cuvette. Combined with 3D mixing channels and matched indicators, GBDT and logarithmic RGB features realize stable pH detection (4.0–9.0) for liquid volumes as low as 1.9 μL. Models are trained with 1890 measurements obtained from 6 replicate measurements of 265 clinical urine samples and 50 simulated samples. In random testing, 92.5% of samples have residuals below 0.2 pH units, and all samples fell within the total allowable error (TEa) of 0.5 pH units for the dipstick method. The maximum SD obtained from the measurements was 0.066 in the 48 h continuous stability test. Under the experimentally tested conditions, the system tolerated urine yellow background up to G4, turbidity up to 200 NTU, mild hematuria/hemolysis up to R3, and variations in ionic strength down to 10 −6 . This method demonstrates potential for certain clinical urine pH testing applications and could be applicable to trace pH measurement in selected life science scenarios, including cell culture, organoid culture, continuous-flow microfluidics, and POCT. Download: Download high-res image (174KB) Download: Download full-size image The pH monitoring system for trace urine analysis in microfluidic chips