Biology subjects

Eskofier, B.

Publications and source records attributed to Eskofier, B..

7 recordsLinked to original sources

bioRxiv2024

Investigating cerebellar control in slow gait adaptations: Insights from predictive simulations of split-belt walking

BackgroundDuring split-belt treadmill walking, neurotypical humans exhibit adaptations characterized by a gradual decrease in step length asymmetry (SLA) toward or beyond symmetry, whereas individuals with cerebellar damage do not show these motor adaptations. Neuromusculoskeletal simulations may help to better understand individual aspects of the underlying neural control, but are currently incapable of predicting adaptations to the continuous perturbations imposed by split-belt walking. MethodsWe extend a spinal reflex model with a biologically inspired model of the cerebellum, which enables error-based motor adaptation by modulating spinal control parameters in response to mismatches between a continuously updated internal prediction and actual motor outcomes. In this work, the cerebellum modulates only a single spinal control parameter, the timing of swing initiation in each leg, which allows examining its isolated contribution to gait adaptation as all other reflex pathways are held constant. We created 80 s predictive simulations of the model walking on a split-belt treadmill with a 2:1 belt-speed ratio, and compared predicted spatiotemporal parameters and kinematics to the reflex-only model and literature. ResultsThe reflex-only model could walk on the split-belt treadmill, but showed no step length adaptations. In contrast, the extended model adapted SLA from an initial asymmetric value toward symmetry or beyond, following an exponential time course similar to that observed in experiments. The model could adapt at varying rates and converge to different asymmetry levels. We found that, in simulation, SLA adaptation during split-belt walking is possible without changes in reflex gains, by adapting the timing of swing initiation. The modulation of timing alone also predicted the experimentally observed exponential adaptation in the temporal domain, but only a linear change in the spatial domain, indicating that additional control mechanisms are likely required to reproduce the full spatial adaptation observed in split-belt walking. ConclusionWe propose a computational model of the cerebellum which, when integrated into a spinal reflex model, autonomously drives feedforward gait adaptations during split-belt walking. This advances the current state of predictive simulations and may eventually help to better understand specific adaptation processes. The modular framework can be extended to test different hypotheses about motor control and adaptation during continuous perturbation tasks.

neuroscience↗

bioRxiv2023

Comparing sparse inertial sensor setups for sagittal-plane walking and running reconstructions

Estimating spatiotemporal, kinematic, and kinetic movement variables with little obtrusion to the user is critical for clinical and sports applications. Previously, we developed an approach to estimate these variables from measurements with seven lower-body inertial sensors, i.e., the full setup, using optimal control simulations. Here, we investigated if this approach is similarly accurate when using sparse sensor setups with less inertial sensors. To estimate the movement variables, we solved optimal control problems on sagittal plane lower-body musculoskeletal models, in which an objective was optimized that combined tracking of accelerometer and gyroscope data with minimizing muscular effort. We created simulations for 10 participants at three walking and three running speeds, using seven sensor setups with between two and seven sensors located at the feet, shank, thighs, and/or pelvis. We calculated the correlation and root mean square deviations (RMSDs) between the estimated movement variables and those from inverse analysis using optical motion capture (OMC) and force plate data. We found that correlations between IMU- and OMC-based variables were high for all sensor setups, while including all sensors did not necessarily lead to the smallest RMSDs. Setups without a pelvis sensor led to too much forward trunk lean and inaccurate spatiotemporal variables. RMSDs were highest for the setup with two foot-worn IMUs. The smallest setup that estimated joint angles as accurately as the full setup (<1 degree difference in RMSD) was the setup with IMUs at the feet and thighs. The mean correlations for joint angles, moments, and ground reaction forces were at least 0.8 for walking and 0.9 for running when either a pelvic sensor or thigh sensors were included. Therefore, we conclude that we can accurately perform a comprehensive sagittal-plane motion analysis with sparse sensor setups when sensors are placed on the feet and on either the pelvis or the thighs.

bioengineering↗

bioRxiv2022

Non-invasive muscle biopsy: estimation of muscle fibre size from a neuromuscular interface

Because of the biophysical relation between muscle fibre diameter and the propagation velocity of action potentials along the muscle fibres, motor unit conduction velocity (MUCV) could be a non-invasive index of muscle fibre size in humans. However, the relation between MUCV and fibre size has been only assessed indirectly in animal models and in human patients with invasive intramuscular EMG recordings, or it has been mathematically derived from computer simulations. By combining advanced non-invasive techniques to record motor unit activity in vivo, i.e., high-density surface EMG, with the gold standard technique for muscle tissue sampling, i.e., muscle biopsy, here we investigated the relation between the conduction velocity of populations of motor units identified from the biceps brachii muscle, and muscle fibre diameter. Moreover, we demonstrate the possibility to predict muscle fibre diameter (R2 = 0.66) and cross-sectional area (R2 = 0.65) from conduction velocity estimates with low systematic bias (~2% and ~4% respectively) and a relatively low margin of individual error (~8% and ~16%, respectively). The proposed neuromuscular interface opens new perspectives in the use of high-density EMG as a non-invasive tool to estimate muscle fibre size without the need of surgical biopsy sampling. The non-invasive nature of high-density surface EMG for the assessment of muscle fibre size may be useful in studies monitoring child development, aging, space and exercise physiology. SIGNIFICANCE STATEMENTOur study explored the relation between the conduction velocity of populations of motor units and muscle fibre size in healthy humans. Our results provide in vivo evidence that a high-density surface EMG-derived physiological parameter, i.e. motor unit conduction velocity, can be adopted to estimate muscle fibre size, without the need of surgical biopsy sampling. Here we propose a neuromuscular interface that opens new perspectives not only in the study of neuromuscular disorders, but also in other fields where the non-invasive and painless determination of muscle fibre and motor unit size becomes a priority, such as in aging, space and exercise physiology.

physiology↗

bioRxiv2022

Sensing the Full Dynamics of the Human Hand with a Neural Interface and Deep Learning

Theories about the neural control of movement are largely based on movement-sensing devices that capture the dynamics of predefined anatomical landmarks. However, neuromuscular interfaces such as surface electromyography (sEMG) can potentially overcome the limitations of these technologies by directly sensing the motor commands transmitted to the muscles. This allows for the continuous, real-time prediction of kinematics and kinetics without being limited by the biological and physical constraints that affect motion-based technologies. In this work, we present a deep learning method that can decode and map the electrophysiological activity of the forearm muscles into movements of the human hand. We recorded the kinematics and kinetics of the human hand during a wide range of grasping and individual digit movements covering more than 20 degrees of freedom of the hand at slow (0.5 Hz) and fast (1.5 Hz) movement speeds in healthy participants. The input of the model consists of three-hundred EMG sensors placed only on the extrinsic hand muscles. We demonstrate that our neural network can accurately predict the kinematics and contact forces of the hand even during unseen movements and with simulated real-time resolution. By examining the latent space of the network, we find evidence that it has learned the underlying anatomical and neural features of the sEMG that drive all hand motor behaviours.

neuroscience↗

bioRxiv2022

Neural decoding from surface high-density EMG signals: influence of anatomy and synchronization on the number of identified motor units

ObjectiveHigh-density surface electromyography (HD-sEMG) allows the reliable identification of individual motor unit (MU) action potentials. Despite the accuracy in decomposition, there is a large variability in the number of identified MUs across individuals and exerted forces. Here we present a systematic investigation of the anatomical and neural factors that determine this variability. ApproachWe investigated factors of influence on HD-sEMG decomposition, such as synchronization of MU discharges, distribution of MU territories, muscle-electrode distance (MED - subcutaneous fat thickness), maximum anatomical cross-sectional area (ACSAmax), and fiber CSA. For this purpose, we recorded HD-sEMG signals, ultrasound, magnetic resonance imaging, and muscle biopsy of the biceps brachii muscle from two groups of participants - untrained-controls (UT=14) and strength-trained (>3 years of training, ST=16) - while they performed isometric ramp contractions with elbow flexors (at 15, 35, 50 and 70% maximum voluntary torque - MVT). We assessed the correlation between the number of accurately detected MUs by HD-sEMG decomposition and each measured parameter, for each target force level. Multiple regression analysis was then applied. Main resultsST subjects showed lower MED (UT: 4.8 {+/-} 1.4 vs. ST: 3.7 {+/-} 0.8 mm) associated to a greater number of identified motor units (UT: 21.3 {+/-} 10.2 vs. ST: 29.2 {+/-} 11.8 MUs/subject). Both groups showed a negative correlation between MED and the number of identified MUs at low forces (r= -0.6, p=0.002 at 15% MVT). Moreover, the number of identified MUs was positively correlated to the distribution of MU territories (r=0.56, p=0.01) and ACSAmax (r=0.48, p=0.03) at 15% MVT. By accounting for all anatomical parameters, we were able to partly predict the number of decomposed MUs at low but not at high forces. SignificanceOur results confirmed the influence of subcutaneous tissue on the quality of HD-sEMG signals and demonstrated that MU spatial distribution and ACSAmax are also relevant parameters of influence for current decomposition algorithms.

bioengineering↗

bioRxiv2022

Common synaptic inputs are not distributed homogeneously among the motor neurons that innervate synergistic muscles

The force generated by the muscles involved in an action is produced by common synaptic inputs received by the engaged motor neurons. The purpose of our study was to identify the low-dimensional latent components, defined hereafter as neural modules, underlying the discharge rates of the motor units from two knee extensors (vastus medialis and lateralis) and two hand muscles (index and thumb muscles) during isometric contractions. The neural modules were extracted by factor analysis from the pooled motor units and no assumptions were made regarding the orthogonality of the modules or the association between the modules and each muscle. Factor analysis identified two independent neural modules that captured most of the covariance in the discharge rates of the motor units in the synergistic muscles. Although the neural modules were strongly correlated with the discharge rates of motor units in each of the synergistic pair of muscles, not all motor units in a muscle were correlated with the neural module for that muscle. The distribution of motor units across the pair of neural modules differed for each muscle: 80% of the motor units in first dorsal interosseous were more strongly correlated with the neural module for that muscle, whereas the proportion was 70%, 60%, and 45% for the thenar, vastus medialis, and vastus lateralis muscles. All other motor units either belonged to both modules or to the module for the other muscle (15% for vastus lateralis). Based on a simulation of 480 integrate-and-fire neurons receiving independent and common inputs, we demonstrate that factor analysis identifies the three neural modules with high levels of accuracy. Our results indicate that the correlated discharge rates of motor units arise from at least two sources of common synaptic input that are not distributed homogeneously among the motor neurons innervating synergistic muscles.

neuroscience↗

bioRxiv2021

Heart rate variability predicts decline in sensorimotor rhythm control

Voluntary control of sensorimotor rhythms (SMR, 8-12 Hz) can be used for brain-computer interface (BCI)-based operation of an assistive hand exoskeleton, e.g., in finger paralysis after stroke. To gain SMR control, stroke survivors are usually instructed to engage in motor imagery (MI) or to attempt moving the paralyzed fingers resulting in task- or event-related desynchronization (ERD) of SMR (SMR-ERD). However, as these tasks are cognitively demanding, especially for stroke survivors suffering from cognitive impairments, BCI control performance can deteriorate considerably over time. It would thus be important to identify biomarkers that predict decline in BCI control performance within an ongoing session in order to optimize the man-machine interaction scheme. Here we determine the link between BCI control performance over time and heart rate variability (HRV). Specifically, we investigated whether HRV can be used as a biomarker to predict decline during voluntary control of SMR-ERD across 17 healthy participants using Granger causality. SMR-ERD was visually displayed on a screen. Participants were instructed to engage in MI-based SMR-ERD control over two consecutive runs of 8.5 minutes each. During the second run, task difficulty was gradually increased. While control performance (p = .18) and HRV (p = .16) remained unchanged across participants during the first run, during the second run, both measures declined over time at high correlation (performance: -0.61%/10s, p = 0; HRV: -0.007ms/10s, p < .001). We found that HRV Granger-caused BCI control performance (p < .001) exhibited predictive characteristics of HRV on an individual participant level. These results suggest that HRV can predict decline in BCI performance paving the way for adaptive BCI control paradigms, e.g., to individualize and optimize assistive BCI systems in stroke.

neuroscience↗