A new optical flow model for motor unit conduction velocity estimation in multichannel surface EMG

• Published in: Computers in biology and medicine Vol. 83; pp. 59 - 68
• Main Authors: Potočnik, Božidar, Holobar, Aleš
• Format: Journal Article
• Language: English
• Published: United States Elsevier Ltd 01.04.2017; Elsevier Limited
• Subjects: Acquisitions & mergers; Action potential; Algorithms; Alignment; Arrays; Biological evolution; Central nervous system; Compensation; Computation; Computer applications; Computer Simulation; Conductors; Electrodes; Electromyography; Estimation; Excitability; Fatigue; Filters; High resolution; Histocompatibility antigen HLA; Humans; Image processing; Internal Medicine; Learning algorithms; Mathematical models; Methods; Misalignment; Models, Neurological; Modified optical flow model; Motor neurons; Motor Neurons - physiology; Motor task performance; Muscle contraction; Muscle Contraction - physiology; Muscle fibre conduction velocity estimation; Muscle, Skeletal - physiology; Muscles; Nervous system; Neural Conduction - physiology; Neuromuscular Junction - physiology; Neuromuscular junctions; Other; Physiology; Reproducibility of Results; Robustness (mathematics); Sensitivity and Specificity; Separation; Signal processing; Single-differential (bipolar) EMG acquisition; Skeletal muscle; Skin; Spatial filtering; Spatial resolution; Surface electromyography; Synthetic EMG signals; Temperature effects; Tracking; Two dimensional flow; Velocity; Single-differential (bipolar) EMG acquisition; Synthetic EMG signals; Surface electromyography; Modified optical flow model; Muscle fibre conduction velocity estimation
• ISSN: 0010-4825; 1879-0534; 1879-0534
• DOI: 10.1016/j.compbiomed.2017.02.006

Many studies have demonstrated the feasibility and benefits of Conduction Velocity (CV) estimation from surface electromyograms (EMGs) in various experimental conditions. Among them, a method based on optical flow was proposed recently, demonstrating relatively accurate CV estimation for EMG signals acquired in monopolar mode. We extended this method by a new data model that compensates more realistically for the spatial Motor Unit Action Potential (MUAP) shape variability and enables accurate CV estimation also in single-differential acquisition mode. The proposed modification was validated on 5000 synthetic Motor Units (MUs) with known CV and direction of fibres. It was shown that, in the noiseless case, the mean CV estimation error was significantly lower for our proposed modification compared to the original CV estimation procedure by up to 2% in the case of monopolar EMG signals and by up to 18.6% for single-differential EMG signals. When estimating fibre directions, the mean error was lower by up to 2.4° (for monopolar EMG signals) and 9.6° (for single-differential EMG signals). The results of tests with 10dB and 20dB noise further demonstrated the robustness of the proposed algorithm to noise in MUAP estimation. •An optical flow data model for motor unit conduction velocity estimation is proposed.•This model is developed for high-density surface electromyograms of fusiform muscles.•Conduction velocity estimation is not sensitive to the orientation of fibres.•Model for motor unit action potential maps was validated by Lucas-Kanade approach.•Significantly lower estimation errors were obtained compared to the state-of-the-art.