Abstract
Rapid and actionable pathogen identification remains a major unmet need in the diagnosis of prosthetic joint infection (PJI). Current diagnostic approaches either provide rapid host response information without pathogen specificity or identify pathogens with delays of days to weeks. Here, we report a chemiresistive nanosensor array combined with machine learning analysis for same-day, pathogen-specific detection based on volatile organic compound (VOC) profiling. A 19-channel nanosensor array was first validated in vitro against a panel of ESKAPEE pathogens, achieving 96% mean classification accuracy using a radial-basis-function support vector machine (SVM) classifier. Data-driven optimization yielded a reduced six-sensor array with high signal-to-noise performance. The optimized platform was evaluated using pooled, uninfected human synovial fluid enriched 1:1 with nutrient media and spiked with Staphylococcus aureus, Staphylococcus epidermidis, or Pseudomonas aeruginosa across a range of 1-10(6) CFU/mL. All infected samples were detected within 9 h, with distinct VOC signatures enabling accurate pathogen differentiation. Time-to-detection (TTD) demonstrated a strong inverse correlation with initial bacterial concentration, supporting semi-quantitative estimation of bacterial load. Negative controls remained at baseline throughout testing. This chemiresistive VOC-based biosensor platform demonstrates the potential to deliver rapid, integrated detection, identification, and burden estimation of metabolically active PJI pathogens, highlighting its promise for future point-of-care diagnostic applications.