Quantitative Detection of Electrical Impedance Muscle Lactate Based on Residual Graph Attention Network
Tianyu Jiang, Anqi Li, Qiang Zhou, Tao Liu, Bo Sun, Kai Liu, Jiafeng Yao · IEEE Sensors Journal · 2025
This study proposes a novel electrical impedance spectroscopy tomography (EIST) method based on a Residual Graph Attention Network (ResGAT) for the visualization and quantitative assessment of lactate accumulation in muscle tissue. By integrating electrical impedance tomography (EIT) method with Electrical impedance spectroscopy (EIS) method, this approach leverages EIS method to detect and quantify variations in lactate concentration, thereby enabling accurate, non-invasive evaluation of muscle lactate levels. The feasibility of the proposed EIST framework was validated through numerical simulations, which demonstrated that ResGAT effectively localizes lactate targets at various positions within the gastrocnemius muscle, achieving an average intraclass correlation coefficient (ICC) exceeding 0.864. Additionally, EIS-based impedance spectra collected under different lactate accumulation conditions yielded a classification accuracy of up to 0.97 using a Support Vector Machine (SVM). For experimental validation, a leg-shaped agar phantom was constructed, in which ResGAT successfully identified cotton targets soaked with 1 ml of pure lactic acid, achieving an average ICC exceeding 0.7. EIS measurements were performed by sequentially injecting equal volumes of lactic acid (0-2.5 ml) into a fixed location of the phantom. The results showed a corresponding decrease in relaxation impedance Z"relax, confirming the capacity of EIS to quantitatively track lactic concentration. Overall, the proposed method provides a reliable and timely tool for non-invasive monitoring of muscle fatigue based on lactate accumulation.