Description of the steps involved in the forward prediction and validation of the TA muscle force during isometric contractions using the developed subject-specific MN-driven neuromuscular model.

During trapezoidal isometric ankle dorsiflexions (A), HDEMG signals were recorded from the TA muscle with a grid of 256 EMG electrodes, built with four grids of 64 electrodes (B), which was artificially downsampled into two grids of lower electrode density. Using convolutive blind-source separation, the HDEMG signals were decomposed into Nr identified motor unit spike trains (C), from which the discharge activity of the complete MU pool was inferred and reconstructed (D) [41]. The experimental and reconstructed populations of spike trains spi(t) were validated for their accuracy in estimating the normalized effective neural drive to muscle (E). The MN-driven neuromuscular model developed in this study (F), consists of a collection of in-parallel Hill-type MU Force Generators (FGs), each including a Neuromechanical Element (NE) and a Contractile Element (CE), that collectively generate the whole TA muscle force FM. FM was predicted (G) from the experimental spi(t) trains in a blind approach (blue dotted lines), after location of the identified MUs into the MU pool (green dashed lines), and from the completely reconstructed population of spike trains (solid red lines). The neuromuscular model was scaled with subject-specific muscle-tendon properties derived from a subject-specific MSK model (I), obtained from segmented MRI scans (H). The predicted TA force , or , depending on the type of neural control, was finally validated (G) against the experimental TA force FTA (J). FTA was estimated from the recorded ankle torque T and the estimated co-activity of the TA’s agonist and antagonist muscles crossing the ankle joint.