Prediction of Ankle Dorsiflexion Moment by Combined Ultrasound Sonography and Electromyography

A Dual-Modal Approach Using Electromyography and Sonomyography Improves Prediction of Dynamic Ankle Movement: A Case Study

“…A recent review study on interpreting hand motions [35] summarized multiple sensing technologies (like EMG and near-infrared spectroscopy, electroencephalography and electroneurography), known as multi-modal bio-signals, to get a higher limb motion estimation or prediction accuracy than uni-modal bio-signals. Our previous studies also showed a dual-modal approach that combines neuromuscular features from TA's sEMG signals and US imaging [10], [11] improves isometric dorsiflexion moment prediction. The dual approach to combine sEMG signals and US imaging may be beneficial as it captures complementary information of muscle activation levels (electrical aspect) and muscle contractility (mechanical aspect) of the targeted skeletal muscle.…”

“…Surface electromyography (sEMG) uses non-invasive electrodes, attached to the skin, to measure electrical potentials due to the excitation of superficial muscles motor units. sEMG signals' amplitude and frequency positively relate to muscle activation levels [10], [11]. Even in severely weak muscles, it can read the intent despite no observable limb motion [12]- [15].…”

“…Compared to the A-mode US signal, the B-mode US imaging-based sensing interface has the advantages of a high signal-to-noise ratio and directly visualizing morphology deformation. Studies in [10], [26]- [29] extracted a variety of morphological and functional features of B-mode US imaging including, pennation angle (PA), fascicle length (FL), muscle thickness, cross-sectional area, echogenicity, tissue displacement or strain, and tissue optical flow, to represent the muscle deformation changes and predict human joints' postures or volitional motion.…”

“…The current study is a next logical step from our previous study on isometric ankle motion and brings us closer to ultimately address the practical need to assist ankle joint dorsiflexion for patients with drop-foot. Although the HNM that combines sEMG and US imaging as mentioned in [10], [11] is also implementable in a dynamic scenario, a system identification of muscletendon-unit geometry, muscle contraction and activation dynamics, and joint-segment dynamics is needed to realize the joint motion prediction from the neuromuscular signals. Further, the dual-modal approach in [10] determined optimal allocation ratios to combine sEMG and US imaging signals.…”

Ankle Angle Prediction Using a Footwear Pressure Sensor and a Machine Learning Technique

“…Multiple IMUs have proven useful [ 10 , 11 , 12 , 13 ]; however, they are not feasible due to the complex setup for daily use, mechanical instability in installation and high cost. MRI (magnetic resonance imaging) [ 14 ], CT (computer tomography), ultrasound sonography [ 15 ] and X-rays [ 16 ] have been used to generate medical images to measure ankle angles but support only a single image at a stationary position. A SVR (support vector regression) machine learning technique was utilized to a high precision with a pre-recorded data analysis to find hip and knee joint motion data, which were then inserted into a predictive model [ 17 ].…”