Prediction of Ankle Dorsiflexion Moment by Combined Ultrasound Sonography and Electromyography

To provide an effective and safe therapy to persons with neurological impairments, accurate determination of their residual volitional ability is required. However, accurate measurement of the volitional ability, through non-invasive means (e.g., electromyography), is challenging due to signal interference from neighboring muscles or stimulation artifacts caused by functional electrical stimulation (FES). In this work, a new model-based intention detection method that combines signals from both surface electromyography (sEMG) and ultrasound (US) sonography to predict isometric volitional ankle dorsiflexion moment is proposed. The work is motivated by the fact that the US-derived signals, unlike sEMG, provide direct visualization of the muscle activity, and hence may enhance the prediction accuracy of the volitional ability, when combined with sEMG. The weighted summation of sEMG and US imaging signals, measured on the tibialis anterior muscle, is utilized as an input to a modified Hill-type neuromusculoskeletal model that predicts the ankle dorsiflexion moment. The effectiveness of the proposed model-based moment prediction method is validated by comparing the predicted and the measured ankle joint moments. The new modeling method has a better prediction accuracy compared to a prediction model that uses sole sEMG or sole US sonography. This finding provides a more accurate approach to detect movement intent in the lower limbs. The approach can be potentially beneficial for the development of US sonography-based robotic or FES-assisted rehabilitation devices.