Exercise-related physiological state classification via a pattern-driven flexible impedimetric e-tongue using holistic sweat analysis

Abstract Sweat analysis has significant potential in personalized health monitoring and exercise physiology assessment due to its non-invasive and real-time nature. However, the complexity of sweat composition limits the applicability of existing potentiometric or voltammetric single sensors and e-tongue systems in complex analysis, as they rely on the selective detection of a single target analyte using specific electrodes, followed by pattern recognition based on that information. To address this limitation, this study proposes a flexible impedimetric e-tongue system based on a sensor array and machine learning-based pattern recognition for the holistic assessment of artificial sweat, enabling the high-accuracy classification of six exercise-related physiological states. The system is constructed using a laser-induced graphene (LIG) interdigitated electrode array, integrated with multi-walled carbon nanotubes (MWCNTs) and metal oxide nanoparticles (MONPs) composite sensing films, ensuring high sensitivity and stability. Experimental results demonstrate that different sensing elements exhibit complementary responses to major sweat analytes (Na⁺, K⁺, lactic acid, and glucose), forming distinctive sweat fingerprint signals. Key parameters were extracted from impedance spectroscopy (IS) and further analyzed using principal component analysis (PCA) for features transformation and support vector machine (SVM) for classification, achieving an overall 97.69% classification accuracy in identifying exercise-related physiological states. Furthermore, the sensor array exhibited high sensitivity, a broad linear range, excellent repeatability (RSD ≤ 1.67%), reproducibility (RSD ≤ 2.9%), and stability over ~ 20 days, confirming its feasibility for sweat analysis. Compared to potentiometric and voltammetric single sensors and e-tongue systems, the proposed impedimetric e-tongue system does not rely on the selectivity of individual sensors but instead analyzes the overall impedance pattern, significantly enhancing its capability for complex biofluid analysis. In conclusion, this study presents a novel impedimetric e-tongue system as a potential solution for exercise monitoring, health assessment, and personalized medicine, providing new insights into the development of non-invasive sweat analysis technologies.