Smart microfluidic biosensor with polyamine saccharide-GluOx-modified LIG electrodes and real-time ML signal deconvolution for glucose monitoring.

This study reports a highly sensitive, flexible, and intelligent microfluidic glucose biosensor integrating machine learning (ML)-optimized and laser-induced graphene (LIG) electrodes with electrodeposited polyamine saccharide-functionalized glucose oxidase (CS/GluOx) crosslinked with polyethene glycol (PEG). LIG fabrication parameters were optimized using a supervised ML model (random forest regression), achieving R2 = 0.92 and RMSE = 0.14 mC/cm2, while reducing fabrication time by 60 % and improving reproducibility. The CS/GluOx-PEG architecture enhanced catalytic efficiency and mechanical stability, maintaining >90 % of the signal after 250 CV cycles at 45° bending. Electrochemical measurements demonstrated linear glucose detection in the range 0.4-10 mM, with sensitivities of 18.0 μA (mM cm2)-1 (flat) and 17.6 μA (mM cm2)-1 (bent). Additionally, ML-based signal deconvolution enabled simultaneous real-time quantification of glucose and lactate in mixed samples, with RMSE values of 0.18 mM and 0.21 mM, respectively. Compared to current glucose sensors, this platform offers superior electrochemical performance, robustness and intelligent analyte discrimination, representing great potential for wearable CGM systems to improve diabetes management.