Model-based deconvolution of a force signal to estimate motor unit twitch parameters under low-force isometric contractions
Muscle force generation and human movement are organised by the central nervous system and executed by the peripheral nervous system and the muscle fibres through molecular and electrical mechanisms. Over the last half-century, attempts have been made to elucidate these mechanisms in vivo, primarily focusing on the motor unit (MU) activity because of its role as the smallest voluntarily contractible unit. Although it is firmly established that the nervous system controls muscle force by modulating MU activity, it is yet possible to distinguish between the activities of slow- and fast-twitch MUs non-invasively, which is important for rehabilitation and diagnostic purposes. Although different methods exist to extract MU twitch parameters from a force signal, no method can accurately identify a single MU twitch given a single MU spike train. We addressed this problem by developing a model-based deconvolution method. We evaluated the method using a MU-based recruitment model under isometric contractions and tested it on experimental data. We found that the deconvolution method can provide non-biased average twitch parameter estimates with low variance for the latest recruited MUs, irrespective of contraction level. It can estimate average twitch parameters when the underlying MUs comprise unequal successive twitch profiles, the force signal has lower signal-to-noise ratios, or when the spike train includes missed firings at the cost of slightly increased bias or variance. Finally, the method provides twitch parameter estimates that align with the expected MU recruitment characteristics in experimental conditions. To conclude, the deconvolution method may be used to study slow and fast MUs for rehabilitation and neuromuscular diagnostics. Author SummaryTo generate force voluntarily with a specific muscle, the brain plans and sends signals through the spinal cord via motor neurons, each of which communicates with a set of muscle fibres. Together, these muscle fibres and the motor neuron are called a motor unit. In the literature, the neural signals have received much attention, whereas the mechanical force-generating muscle fibres have received much less due to the limitations of current methods. By extracting the mechanical characteristics of these muscle fibres connected to a specific motor neuron type in vivo, one can use this information for rehabilitation and neuromuscular diagnostics of humans. Here, we proposed a method that can accurately estimate the force profile from each motor unit during low to high contraction levels. This method can be used for rehabilitation and neuromuscular diagnostics purposes.