< Back to previous page

Project

A blended experimental and computational approach to assess control of standing balance

With age, our ability to complete seemingly simple activities of daily living declines. Older adults move slower, less accurate and more often lose their balance resulting in frequent falls, which might lead to significant injuries requiring medical attention and leading to loss of independence. Age is associated with declines in muscle strength, sensory acuity, motor acuity and cognitive ability, but our understanding of the relative contribution of these different neural and muscular changes in sensorimotor function to reduced balance control is incomplete. This hampers the design and selection of prevention and rehabilitation therapies and devices that aim to maintain or restore efficient balance control in order to prevent falls.

The combination of experimental data and model-based simulations has much potential to advance our understanding of the mechanisms underlying balance control. Such blended experimental-computational approach has mainly been applied to study the relation between muscular and skeletal properties and movement, but has been used less to study how limitations in the central nervous system affect human movement coordination. The efficient and accurate integration of sensory feedback signals by the central nervous system is key to perform skillful movement but is always imperfect and seems to deteriorate as we age.

The main aim of this thesis was to identify how different age-related changes in the neuro-musculoskeletal system affect reactive balance performance, and how resistance and perturbation training, an established and novel intervention respectively, improve balance control. This overall goal was tackled using a blended experimental-computational approach. Experiments involved a protocol, quantifying balance performance, strength, sensory acuity and motor acuity in both young and older adults. This allowed testing associations between sensorimotor function and reactive balance performance. Experiments were complemented with optimal control simulations based on neuro-musculoskeletal models and were applied to test whether the experimentally established associations might be causal. Four studies were executed to accomplish the overall goal.

In study 1, we developed a new simulation framework that extends currently available frameworks to incorporate sensory noise, motor noise and other types of uncertainty that affect human movement. The presented framework is novel compared to prior simulation approaches as it allows for simulations of systems that are both stochastic and nonlinear. Such an approach was previously only available for simple musculoskeletal models that neglected important features of muscle physiology and nonlinearities in the skeletal dynamics because of limitations in numerical efficiency. Incorporating uncertainty in human movement simulations is important to accurately study the neural control of movement and the limitations related to balance control.

Our new approach combines recent advances in deterministic simulations of human movement with an approximate reformulation of the stochastic simulation problem into a deterministic form. We successfully applied the new framework to generate more accurate and detailed predictions of both standing balance and goal-directed reaching, which we chose as paradigm movement tasks since they are widely studied. We demonstrated that both accurate musculoskeletal representations and physiological noise contribute to the realism of movement predictions. The presented stochastic optimal control framework allowed simulating more realistic movement patterns and neural control laws compared to previously available simulation approaches. Neural control was more accurately simulated by (1) accounting for uncertainty with Gaussian properties, to represent sensory and motor noise or unpredictable interactions with the simulated environment, (2) allowing contributions from both sensorimotor feedback and feedforward controls to movement and (3) enabling the identification of minimum effort solutions under explicit movement accuracy and stability constraints. By increasing the fidelity of neuro-musculoskeletal simulations, we predicted and investigated features of human movement that have been difficult or impossible to study with available simple stochastic models or simulations based on complex deterministic models.

In study 2, we aimed to understand the origins of trial-by-trial (intra-subject) and between-subject (inter-subject) variability in reactive balance responses to platform translations during standing. Reactive balance responses exhibit much intra- and inter-subject variability, which is not well understood. Such variability complicates studying the effect of aging and interventions. A better understanding of intra- and inter-subject variability in healthy young adults will allow better differentiation between groups. We collected the responses of ten young healthy adults to unpredictable platform translations and performed simulations to test for potential causality.

Under the instruction to try and keep the feet in place, the responses to unpredictable platform translations exhibited a continuum of center of pressure strategies (COP strategies), i.e. single pendulum-like rotation around the ankle joint, hip strategies, i.e. counter rotation of leg and trunk segments, and stepping strategies in response to the perturbations. First, our experiments and simulations confirmed that initial posture explains the experimentally observed trial-by-trial variability with a more anterior initial COM position increasing the use of the hip strategy. Second, differences in task-level goal explain observed inter-subject variability with prioritizing effort minimization leading to COP strategies and prioritizing stability leading to hip strategies. Third, interactions between initial posture and task-level goal explained observed differences in intra-subject variability across subjects. Based on these results we formulated outcome measures that normalized for trial-by-trial differences in initial posture, and in turn allowed for better differentiation between groups in later studies.

In study 3, we compared the effectiveness of perturbation and resistance training for reactive balance performance during standing and analyzed the mechanisms through which such training could improve reactive balance performance in different movement tasks. Based on the experimentally determined associations between strength and reactive balance performance, resistance training has been a popular fall prevention therapy. However, such interventions have at most resulted in limited improvements in balance performance. Perturbation training is a more recent training paradigm with some promising results but has not been compared to resistance training in terms of effectiveness to improve reactive balance performance. In addition, the adaptations in balance-correcting mechanisms that underlie improvements in reactive balance control after perturbation and resistance training are not well understood. Moreover, it is unclear whether perturbation and resistance training induce adaptations in balance control that generalize to movement tasks that were not part of the training. We performed two training interventions with two groups of healthy older adults: a 12-week resistance training intervention and a 3-week perturbation training intervention consisting of support-surface perturbations of standing balance. Reactive balance performance during standing and walking as well as a set of neuro-muscular properties were assessed pre- and post-intervention.

Both perturbation training and resistance training induced training specific improvements. Perturbation training was more effective in reducing step incidence during perturbations of standing than resistance training. Resistance training increased maximal strength of the knee extensors whereas perturbation training did not. Following perturbation training, older adults were able to suppress the initiation of a step at higher deviations from upright equilibrium without adaptation in the application of COP and hip strategies. Improvements in reactive balance during standing induced by perturbation training did not generalize to perturbed walking or dynamic balance performance during narrow-beam walking, indicating the specificity of such training interventions and suggesting that different mechanisms drive the selection of a strategy during both tasks.

In study 4, we sought to identify the cause-effect relations between age-related changes in sensorimotor function and reactive balance performance. Many factors of the neuro-musculoskeletal system change with age but our understanding of the relative contribution of these different neural and muscular changes in sensorimotor function to reduced balance control is incomplete. Here we collected a comprehensive dataset to associate vestibular function, proprioceptive function, visual function, sensory reweighting ability, strength and motor accuracy with balance performance across both young and older adults. Next, we tested whether the detected associations might reflect causal relations by applying the novel simulation framework of study 1.

Older adults stepped more often in response to platform translations during standing compared to younger adults, indicating their decreased balance performance. We found that stepping threshold, but not the reliance on hip or COP strategies was associated with step incidence. The increased step incidence was associated with experimental measures of strength, visual acuity and the sensory organization test preference score, which quantifies the ability to suppress conflicting sensory information. Yet, only the sensory organization test preference score correlated with the stepping threshold. Using stochastic optimal control simulations we estimated parameters describing sensory and motor noise in our neuro-musculoskeletal model, which we used in predictive simulations of reactive balance responses to platform translations. Assuming optimal sensory reweighting, our simulations suggested that older adults can compensate for both increasing visual and vestibular noise to maintain reactive balance performance in response to support-surface translations. In simulation, decreased strength limited the efficiency of the COP strategy, which would induce increased step incidence for the highest perturbation magnitudes, but we did not observe age-related changes in the COP strategy in experiments.  Therefore, it remains unclear whether strength might have contributed to limited reactive balance performance in our group of older adults. Taken together, our blended experimental-computational approach suggest that healthy older adults may be able to adapt to alterations in sensory and motor noise, but that they may lack the ability to rapidly adjust their balance control strategy to a different context. A different context refers to a different movement task (e.g. standing and walking) or a different type of perturbation during the same task (e.g. platform translations vs platform rotations or translations in different directions). Further modeling developments are needed to capture the ability to adjust balance control to different contexts, which was measured here by the sensory organization test preference sub scores.

The blended experimental-computational approach applied in this dissertation generated new insights specific to reactive balance control in aging populations. Although experimental measures of strength and sensory acuity on the one hand and reactive balance performance on the other hand are associated, it seems that healthy older adults may be able to compensate for such decreases in sensorimotor function. We propose that rather abilities that determine whether individuals can flexibly adjust sensorimotor responses to different movement contexts limit reactive balance performance in older adults. We demonstrated that such abilities might be improved following perturbation training, where  an increase in the ability to suppress stepping responses to unpredictable multidirectional perturbations, without changes in the application of COP and hip strategies, led to improved reactive balance performance. Future therapeutic approaches might benefit from increasing focus on applying balance perturbations that challenge balance throughout different contexts and require different motor corrections by for example combining perturbations in different directions or combining balance tasks with another task (dual task training).

The combination of collected data on sensorimotor function with simulations of musculoskeletal models that incorporated the effects of uncertainty is innovative and might be useful to address more unanswered research questions on the limitations of human movement due to constraints in both the neural and musculoskeletal system and to develop interventions and devices to alleviate such limitations.

Date:1 Dec 2016 →  31 Dec 2021
Keywords:musculoskeletal, optimal control, computational biomechanics, balance control, fall prevention
Disciplines:Orthopaedics, Human movement and sports sciences, Rehabilitation sciences
Project type:PhD project