Deep learning-based noninvasive assessment of additively manufactured biomechanical models for sports training simulation

Real-time monitoring of the biochemical reactions of athletes is important in terms of prevention of injuries and maximisation of performance. In this research, an IoT-based wearable biosensing platform of electronic chemical is introduced to monitor the persistent sweat biomarkers in collegiate artistic gymnasts. The flexible sensor patch is able to simultaneously detect lactate (120 mM −20 mM), glucose (0.5–5 mM), sodium (100–150 mM), potassium (2.5–5.5 mM), and cortisol (0.120 0.1–2.0 mcg/dL). The sensitivity of lactate sensor was found to be −12.5 +-0.53 nA/mM −1 and the response time was 82 +-8 s. The sweat lactate (compared to blood lactate) had a significant correlation with blood lactate during validation in 480 training sessions (n = 45 athletes, r = 0.92, 95% CI: 0.880.95, p < 0.001). The accuracy of fatigue prediction of integrated multi-biomarker machine learning analysis was AUC = 0.91. The proposed system allows adjusting the training according to the individual requirements and detecting fatigue in real time as well as identifying potential overtraining at the initial phase. This platform exhibits high levels of analytical soundness and feasibility in the monitoring of sports performances.