Development of a stochastic approach to estimate suboptimal control in an adult population

Humans are capable of extraordinary things thanks to the muscles, degrees of freedom and the central nervous system. Understanding how the brain selects motor control strategies in healthy and clinical populations is paramount to advancing our knowledge and better treat neuromuscular diseases. Historically, neuromuscular modelling approaches can be categorised into three groups: i) the reductionist approach, which splits the nervous and musculoskeletal systems, ii) the explicit modelling of the coupling nervous-musculoskeletal systems, and iii) the stochastic approach, which implement the uncontrolled manifold theory to explore the space of plausible neural solutions. Myobolica belongs to the latter, as it employs Bayesian statistics and a Markov Chain Monte Carlo algorithm to predict numerous plausible muscle control strategies. Unlike existing methods, Myobolica incorporates physiological constraints to ensure realistic solutions. To evaluate its performance, public datasets were used. Myobolica successfully simulated lower limb and shoulder biomechanics, showing a good correlation with experimental joint loads (R2 = 0.92, RMSE = 0.3 BW and R2=0.6, RMSE=0.73 BW, respectively). Following, Myobolica was also applied to simulate four patients with Parkinson’s disease. This explorative assessment suggests its potential for deepening disease mechanisms and predicting progression. Before Myobolica can be used to address specific clinical needs, it is paramount to perform a robust validation. A preparatory sensitivity analysis – key part of the validation – on 48 simulations have been discussed to plan an optimal strategy for this future development. In parallel, some effective methods to perform such analysis and communicate obtained results were evaluated.