Assessment of balance and gait abilities in the clinic, laboratory, and natural environments

Balance is fundamental in our daily life, and a proper postural control is essential to maintain it, achieve it, or restore it. Maintaining balance is a complex sensorimotor task and hence it can be affected by ageing, neurological conditions, as well as many activities in our daily routines which can lead to a loss of balance. Thus, assessing postural control is crucial. In the clinical setting, it can be used to conduct early monitoring and prevention, while, for example, in the work environment it could be adopted to correct posture and avoid injuries or pain due to a prolonged situation of discomfort. The knowledge of the daily implications of postural control on the quality of our daily life drove my PhD project and the contents of this thesis: all the studies included within this work examined postural control in different scenarios, tasks, and populations by adopting a variety of techniques and methods in order to have a clear and global understanding of postural control in all its facets. Indeed, this thesis proposes different setups, protocols, and metrics to comprehensively assess postural control in a quantitative manner overcoming the limitation of the methods traditionally used, which often suffer from high subjectivity, flooring/ceiling effects and poor sensitivity. My project started by investigating changes in postural control while standing. Specifically, my first aim was to characterize and mathematically define age-related changes in both static and dynamic balance tasks. To do so, I analyzed the data of 272 healthy subjects, with an age range spanning from 20 to 90, who underwent a test conducted using a medical robotic device, hunova. The results revealed that the decline of balance abilities with age could be described by a quadratic curve and that, as expected, the rate of age-dependent changes is also influenced by the testing conditions. The knowledge of the age's effect on postural control in a large population, covering the entire adult lifespan, can be used to evaluate the possible onset of balance problems, separating them from a normal decay of the balance ability due to age. After evaluating the effect of age on postural control, I chose to investigate how neurological disorders, such as Parkinson's Disease (PD) and Multiple Sclerosis (MS), affect balance at the early stage of the disease, when postural deficits are not clearly evident yet. Using the same setup and a similar protocol, I assessed balance deficits in PD subjects by comparing 10 PDs and 10 age-matched controls. The results highlighted how the robotic platform hunova is, indeed, capable of detecting postural control deficits in people with Parkinson’s even in case of a low Posture Instability-Gait Disturbance (PIGD) score. Then, I studied the effects of MS by comparing 27 subjects with MS and 17 age-matched controls. Here, the focus was on static standing evaluated by structural parameters that are a novelty in MS but are already studied in healthy subjects and PD. Those parameters are strictly related to the underlining neural control process as derive from the sway-density curve (SDC). Indeed, the SDC represents the instants of time in which the ankle torque is stable and detects the restoring force that compensates for the micro-falls that are typical in the dynamic of unperturbed standing and thus the descending (feed-forward) motor commands generating such force. In this study, I found that MS swayed more, and their feed forward commands were larger in amplitude. A greater posturographic command may be a strategy of MS subjects to counteract their deficit to properly calibrate the inverse model of the unstable inverted pendulum that led to the maintenance of an upright standing position. These two studies on neurological subjects highlighted the importance of instrumented balance assessments that can bring to light deficits that are not otherwise identified. Within this thesis, to keep on in the assessments of balance, I also evaluated postural control while sitting. Specifically, two different setups to quantitatively study trunk control in clinical settings are presented. First, I extensively characterized trunk control in 15 stroke subjects using a functional test. In this study, the kinematic performance and muscular activity of the upper body after stroke have been studied during the frontal and lateral modified functional reach test, with a specific and primary focus on trunk control and the corresponding trunk muscles. Then, I evaluated trunk control using the same robotic device already proposed, hunova, which also allows balance assessments while sitting. Here, I tested 18 healthy subjects to investigate the kinematic and muscular activations of the volitional and the reactive components of trunk control. Subjects, seated on the device, were requested to move the seat platform in a well-defined direction or to adapt to the continuous and predictable perturbations of the device. Subjects improved their performance with practice and learnt how to adapt to the perturbations. In this study, the kinematic and muscular performance was studied in dependence of specific parameters of the device which could be set, such as the velocity of the seat motion or its work-space. These results drove us to adapt the protocol to extensively study trunk control in Spinal Cord Injured (SCI) subjects throughout a complete, but clinically applicable, protocol. Here, I assessed the performance of 10 SCI subjects (4 with a complete lesion and 6 with an incomplete lesion of the spinal cord). In my project I have also studied the role of trunk muscles and trunk control while driving throughout a simulator, and while writing in a school environment. These two studies were conducted to highlight the feasibility of using these setups for quantitative assessments of performance in out-of-clinic scenarios. Lastly, I focused on another sensorimotor task: gait. Within this thesis, gait is studied using approaches that does not require the use of markers to avoid modifying the naturalness of the subject’s movements: the 2D and 3D markerless analysis. The former used a single lateral RGB camera while the latter included 3 RGB cameras. Deep learning algorithms are then used to extract key points from which the subject’s movement was assessed. 16 stroke survivors have been tested with a 2D setup, while 16 healthy subjects underwent an experiment using a 3D setup. Results of both studies were promising. The first study did not highlight significant differences between marker-based and markerless gait analysis: both the spatio-temporal parameters and the elevation angles in the sagittal plane computed with the markerless approach did not differ from the one computed with the marker approach. Also, those parameters detected the differences between the two legs of stroke subjects. Indeed, the second study (i.e., the one in 3D) showed comparable results when considering spatio-temporal parameters and joint angles. The only difference was an underestimation of the maximum flexion for ankle and knee angles. These results highlighted the possibility to adopt markerless technique for gait analysis inside the clinical settings, however, these setups can be easily adapted for gait analysis, and more in general, human motion analysis, outside in a freer environment. Hence, this thesis assesses postural control in terms of standing and sitting balance, but also includes its assessment during other dynamic activities. This thesis proved the usability of these setups for comprehensive and quantitative assessments in different scenarios, and also the use of these to answer scientific questions. Indeed, this thesis provided technologies, protocols and metrics that can be used and/or translate in different scenarios to allow quantitative and standardized assessments of postural control. The results reported within this thesis also contribute to enlarge the knowledge on postural control.