Falling is a problem in elderly, affecting 30% of people over 65 years old, and leading to serious injuries in 10% of the cases. Because balance is controlled in a closed loop system with a high level of redundancy, the cause of impaired balance control is often unknown. A balance control model in combination with perturbed balance experiments could be a useful aid to detect the underlying cause of balance impairment, which would allow for better treatment. A disturbance estimation and compensation (DEC) model has a strong neurological basis and is able to describe multiple conditions with one set of parameters. An existing DEC model, containing separate ankle and hip control, was extended by adding visual and mechanical disturbance estimations. The model was fitted to experimental data of elderly subjects maintaining balance in three conditions containing four disturbances simultaneously; visual and ankle proprioceptive perturbations and mechanical perturbations at the hip and shoulder level. To validate the fitting procedure, the model parameters were also fitted to data simulated by the model, and a sensitivity analysis was done.Being fitted to experimental data, the model could reproduce about 60% of the behaviour. Mainly the low frequency behaviour could not fully be tracked. Parameter estimation proved challenging, with some parameters consistently estimated larger or smaller than expected or physiologically plausible. Fitting the model on its own simulated data could reproduce >99.9% of the behaviour, but with some parameters consistently estimated up to a factor 10 too small or too large. Many adaptations and extensions to the existing DEC model were made simultaneously and with the available experimental data it was not possible to investigate the individual influence of each adaptation. If a broader range of experimental data is available, with conditions where only one, two or three perturbations are applied, it is possible to validate each model adaptation separately. This would likely improve the model. With the suggested improvements, the multisegmental DEC model opens up the way towards better treatment for balance impairment.

Introduction Stroke often initially leads to motor impairments of the upper limb. The largest improvements in recovery take place during the _rst half year after stroke. With more knowledge about neurological recovery, it may be possible to predict motor recovery after stroke with the use of neurological biomarkers. Prediction of stroke recovery can be used in the selection and development of rehabilitation therapies. By applying continuous perturbations to a joint with the use of a robotic manipulator during di_erent control tasks, the evoked cortical responses of the sensorimotor cortex can be studied. With this method, evaluating the relation between EEG activity and motor function of the upper limb may provide meaningful EEG biomarkers for motor recovery after stroke. Methods A selection of 13 sub-acute stroke patients who participated in the 4D EEG project were longitudinally measured four times during the _rst half year after stroke. EEG was measured during relax, position and force tasks while a robotic wrist manipulator applied continuous perturbations to the a_ected wrist. The upper limb part of the Fugl-Meyer assesment (FMA) was used to asses the upper limb motor function. The index of the signal-to- noise ratio (iSNR) was obtained from the EEG data. Results The FMA score was strongly related to time post stroke. Also, FMA score during the _rst measurement was lower than during the second, third and fourth. The iSNR was lower during the position task than during the relax and force tasks. A moderate relation was found between the FMA score and the iSNR during the position task, but not during the relax and force task. Conclusion The iSNR during the position task is potentially a EEG biomarker for upper limb motor recovery after stroke. 1

Motor learning is a vital ability of the human brain in which multiple cortical areas like the primary motor cortex are involved. One way to investigate the fast motor learning brain dynamics is with use of electroencephalography (EEG). This non-invasive and mobile technique records continuous electrical activity on the brain cortex with a high temporal resolution. Although there has been considerable research into motor learning in humans, the mechanisms behind movement acquisition and execution are still largely unknown. A better understanding of motor learning is relevant to treatment and training in neurorehabilitation and sports medicine. The aim of this study was to provide insight into the neurophysiological mechanisms behind motor skill learning compared to motor execution. Motor skill learning is defined as the acquisition of a complex movement sequence by improving accuracy without comprising on speed. To investigate the neurophysiological mechanisms behind motor skill learning, 128-channel EEG was recorded in 20 young, right-handed, healthy participants while performing a motor skill learning task and a motor execution task. With use of adaptive mixture independent component analysis (AMICA), time-frequency analysis and equivalent dipole fitting source localisation, functional neurophysiological correlates of motor skill learning were investigated. Due to an excessive amount of electrical bridged electrodes, 12/20 participants were excluded from further analyses. In the 8 remaining participants, a cluster of independent components of electrical activity was located on the left primary motor cortex. In the motor learning task, lower β-frequency power was found in these components compared to the control task. This suggest that motor execution could be distinguished from motor learning on EEG by means of β activity in the left primary motor cortex. The results of this study can be used for future research on motor skill recovery in rehabilitation, motor learning in sports medicine and research on interventions to enhance motor learning, like non-invasive electrical brain stimulation.