Analysis of Muscle Fatigue Progression Using Geometric Features of Surface Electromyography Signals and Explainable XGBoost Classifier

PurposeAnalysing the progression of muscle fatigue is paramount as it can be helpful in monitoring the myoelectric manifestations of fatigue conditions and predicting any abnormalities at an early stage. In this work, a detection model for tracking the muscle fatigue progression is developed using geometric features of surface electromyography (sEMG) signals and explainable eXtreme Gradient Boosting (XGBoost) classifier.MethodsThe signals are recorded under dynamic contractions from fifty-eight healthy adult volunteers. These signals are preprocessed and divided into three equal zones: nonfatigue, transition-to-fatigue and fatigue. These segments are subjected to Hilbert transform, and the resultant coefficients are represented in a complex plane to form a shape. The geometric features namely, perimeter, area, second moment, inertia and instantaneous spectral centroid (ISC) are extracted from the shape and a classification model is developed using XGBoost and Shapley Additive exPlanations (SHAP) approach.ResultsFour features, viz. perimeter, area, second moment and inertia, follow an increasing trend towards fatigue, whereas ISC decreases towards fatigue. The features facilitate the discrimination of nonfatigue and transition-to-fatigue zones. However, there is an overlap between transition-to-fatigue and fatigue zones. Interestingly, the classification model achieves a balanced accuracy and F-score of 96.83% and 95.25% for differentiating transition-to-fatigue and fatigue zones. SHAP values reveal that the impact of ISC is more for the classification of three zones.ConclusionThe geometric features are able to characterise the sEMG signals during the progression of fatigue. The proposed approach could be helpful to track the muscle fatigue progression in applications such as sports biomechanics, rehabilitation and myoelectric prosthesis.