Convolutional neural network model-based prediction of human muscle activity by analyzing urine in body fluid using Raman spectroscopy.

This article focuses on a novel method for quantitatively analyzing urea concentration in urine using Raman spectroscopy combined with a convolutional neural network (CNN) to assess muscle fatigue. The study developed a database of 2000 Raman spectra from urea solutions of varying concentrations and demonstrated that the CNN model achieved high accuracy (R² ≈ 0.9993) and low error (RMSE ≈ 0.0145) in predicting urea levels. Compared to traditional optical colorimetric methods, this approach offers greater stability, efficiency, and does not require sample pretreatment, making it suitable for rapid, noninvasive monitoring of muscle fatigue in athletes. The method also showed good selectivity and sensitivity despite potential interference from other urine components, supporting its potential application in sports medicine and health monitoring.