Dynamic Muscle State-Driven Framework for Intelligent Grading of Chronic Trapezius Fatigue

Prolonged sedentary work often induces chronic shoulder muscle fatigue, which can lead to persistent pain. Objective quantification of cumulative fatigue severity is critical for effective intervention, yet existing methods lack standardized multidimensional assessment tools. Focusing on the trapezius, a key stabilizer of the shoulder girdle and a common site of pain, this study proposes a novel muscle mechanical state-based grading framework for chronic fatigue. Muscle tone and stiffness were used as dual biomarkers to characterize muscle rigidity. A two-phase experimental paradigm was designed to analyze the temporal evolution and correlation patterns of these two parameters during fatigue progression. Based on the dynamic regulatory capacity reflected in muscle mechanical state data, a multi-level grading system for chronic muscle fatigue was established. For intelligent assessment, three dual-branch graph-structured models were developed under a progressive framework of "independence → simple interaction → deep interaction." The final HG-GCN model employed cross-attention to associate the spatiotemporal features of the middle and lower trapezius and dynamically fused features through a gated mechanism. The model achieved classification accuracies of 88.67% for the middle trapezius and 89.17% for the lower trapezius, demonstrating its effectiveness in detecting and quantifying chronic fatigue for early intervention.