A pH-Calibrated Multi-Analyte MXene Aerogel Electrochemical Platform Assisted by Interpretable Machine Learning for CSF-Based Prescreening of Postoperative Bacterial Meningitis

Rapid and reliable prescreening and early risk stratification of postoperative meningitis (PNM) remain challenging because clinical symptoms are nonspecific and culture-based testing is time-consuming. Here, we report a machine-learning-assisted electrochemical platform for point-of-care (POC) prescreening and triage of PNM using cerebrospinal fluid (CSF). A MoS2/MXene/reduced graphene oxide (rGO)-based multiplex sensor array enables concurrent quantification of pH, glucose, and lactate-three clinically relevant biomarkers for postoperative bacterial meningitis (PNBM). To ensure stable and reproducible multiplex sensing, a hierarchically porous, ant-nest-like (macropore-mesopore) MoS2/MXene/rGO (MMG) aerogel is engineered as the electrochemical transduction scaffold, providing interconnected transport pathways while mitigating 2D nanosheet restacking. In addition, real-time pH monitoring is integrated to correct the intrinsically pH-dependent glucose and lactate responses, improving quantitative consistency across physiologically variable samples. Based on a clinically relevant PNBM/non-PNBM cohort, interpretable tree-ensemble models (Random Forest and XGBoost) are trained using Gaussian-noise augmentation within cross-validation folds to avoid information leakage. The resulting framework demonstrates preliminary internal cross-validated discrimination, while maintaining transparent decision logic. Overall, this work integrates pH-calibrated multiplex electrochemical sensing with interpretable machine learning, offering a rapid and transparent strategy for PNBM prescreening in neurosurgical POC settings.