Stress fusion evaluation modeling and verification based on non-invasive blood glucose biosensors for live fish waterless transportation

Non-invasive blood glucose level (BGL) evaluation technology in skin mucus is a wearable stress-detection means to indicate the health status of live fish for compensating the drawbacks using traditional invasive biochemical inspection. Nevertheless, the commonly used methods cannot accurately obtain the BGL variations owing to the influence of an uncertain glucose exudation rate, ambient effects, and individualized differences. Our study proposes a non-invasive multi-sensor-fusion-based method to evaluate the dynamic BGL variations using the enhanced gray wolf-optimized backpropagation network (EGWO-BP) to continuously acquire more accurate trends. Furthermore, the K-means++ (KMPP) algorithm is utilized to further improve the accuracy of BGL acquisition by clustering fish with full consideration of its size features. In the verification test, turbot (Scophthalmus Maximus) was selected as an experimental subject to perform the continuous BGL monitoring in waterless keep-alive transportation by acquiring comprehensive biomarker information from different parts of fish skin mucus, such as fins, body, and tails. The comparison of results indicates that the KMPP-EGWO-BP can effectively acquire more accurate BGL variation than the traditional gray wolf-optimized backpropagation network (GWO-BP), particle swarm-optimized backpropagation network (PSO-BP), backpropagation network (BP), and support vector regression (SVR) by mean absolute percentage error (MAPE), root mean square error (RMSE), and coefficient of determination ( R ²). Finally, the proposed BGL fusion evaluation model can precisely acquire the live fish's physiological stress states to substantially reduce the potential mortality for the live fish circulation industry.