Dickerson, C. R. | bioRxiv Science

bioRxiv2025

A methodological framework for assessing neural control of shoulder muscles during multi-directional isometric tasks using high-density surface electromyography.

The complex function of the shoulder relies on the coordinated activation of small and large muscles, including the deltoid, pectoralis major, trapezius, and latissimus dorsi. However, detailed knowledge of their neuromuscular control remains limited. This study aimed to develop a methodological framework to investigate the neural control of the larger superficial shoulder muscles by combining a six-degree-of-freedom load cell attached to a robotic arm with high-density surface electromyograms (HDsEMG). Six healthy participants performed isometric contractions (abduction, adduction, flexion, and extension) at 30% of maximal voluntary contraction with the shoulder positioned at 30{degrees} and 65{degrees} of lateral abduction. HDsEMGs were recorded from the four muscles and analysed at the global activation, spatial distribution of activation and motor unit levels. Global activation was quantified using averaged normalized root-mean-square (RMS) amplitude and spatial distribution using coefficient of variation of the topographic maps. Moreover, HDsEMGs were decomposed into individual motor unit spike trains using convolutive blind source separation, and motor unit behaviour was characterized by mean discharge rate and spatial distribution of motor unit action potentials (MUAPs). RMS maps revealed action-specific activation within and between muscles, with the upper trapezius active across all tasks, while the anterior, middle, and posterior deltoid, clavicular pectoralis major, and latissimus dorsi were predominantly activated during abduction, flexion, and extension. Motor unit discharge rate also showed task-dependent activity. MUAP spatial distributions further showed distinct motor unit territories within arrays, suggesting region-specific recruitment strategies across actions. In conclusion, this framework demonstrates that individual motor unit activity can be reliably measured non-invasively in the superficial shoulder muscles. The approach provides a methodological basis for novel incorporation of neural control information into biomechanical models of shoulder function.

neuroscience↗

bioRxiv2021

MuscleNET: Mapping Electromyography to kinematic and dynamic biomechanical variables by machine learning

ObjectiveThis paper proposes machine learning models for mapping surface electromyography (sEMG) signals to regression of joint angle, joint velocity, joint acceleration, joint torque, and activation torque. ApproachThe regression models, collectively known as MuscleNET, take one of four forms: ANN (Forward Artificial Neural Network), RNN (Recurrent Neural Network), CNN (Convolutional Neural Network), and RCNN (Recurrent Convolutional Neural Network). Inspired by conventional biomechanical muscle models, delayed kinematic signals were used along with sEMG signals as the machine learning models input; specifically, the CNN and RCNN were modeled with novel configurations for these input conditions. The models inputs contain either raw or filtered sEMG signals, which allowed evaluation of the filtering capabilities of the models. The models were trained using human experimental data and evaluated with different individual data. Main resultsResults were compared in terms of regression error (using the root-mean-square) and model computation delay. The results indicate that the RNN (with filtered sEMG signals) and RCNN (with raw sEMG signals) models, both with delayed kinematic data, can extract underlying motor control information (such as joint activation torque or joint angle) from sEMG signals in pick-and-place tasks. The CNNs and RCNNs were able to filter raw sEMG signals. SignificanceAll forms of MuscleNET were found to map sEMG signals within 2 ms, fast enough for real-time applications such as the control of exoskeletons or active prostheses. The RNN model with filtered sEMG and delayed kinematic signals is particularly appropriate for applications in musculoskeletal simulation and biomechatronic device control.

systems biology↗