Balance impairment is a major mechanism behind falling along with environmental hazards. Under physiological conditions, ageing leads to a progressive decline in balance control per se. Moreover, various neurological disorders further increase the risk of falls by deteriorating specific nervous system functions contributing to balance. Over the last 15 years, significant advancements in technology have provided wearable solutions for balance evaluation and the management of postural instability in patients with neurological disorders. This narrative review aims to address the topic of balance and wireless sensors in several neurological disorders, including Alzheimer’s disease, Parkinson’s disease, multiple sclerosis, stroke, and other neurodegenerative and acute clinical syndromes. The review discusses the physiological and pathophysiological bases of balance in neurological disorders as well as the traditional and innovative instruments currently available for balance assessment. The technical and clinical perspectives of wearable technologies, as well as current challenges in the field of teleneurology, are also examined.

Ergonomics evaluation through measurements of biomechanical parameters in real time has a great potential in reducing non-fatal occupational injuries, such as work-related musculoskeletal disorders. Assuming a correct posture guarantees the avoidance of high stress on the back and on the lower extremities, while an incorrect posture increases spinal stress. Here, we propose a solution for the recognition of postural patterns through wearable sensors and machine-learning algorithms fed with kinematic data. Twenty-six healthy subjects equipped with eight wireless inertial measurement units (IMUs) performed manual material handling tasks, such as lifting and releasing small loads, with two postural patterns: correctly and incorrectly. Measurements of kinematic parameters, such as the range of motion of lower limb and lumbosacral joints, along with the displacement of the trunk with respect to the pelvis, were estimated from IMU measurements through a biomechanical model. Statistical differences were found for all kinematic parameters between the correct and the incorrect postures (p < 0.01). Moreover, with the weight increase of load in the lifting task, changes in hip and trunk kinematics were observed (p < 0.01). To automatically identify the two postures, a supervised machine-learning algorithm, a support vector machine, was trained, and an accuracy of 99.4% (specificity of 100%) was reached by using the measurements of all kinematic parameters as features. Meanwhile, an accuracy of 76.9% (specificity of 76.9%) was reached by using the measurements of kinematic parameters related to the trunk body segment.

Markerless human motion capture could be a potential tool for studying body biomechanics. Despite the extensive previous studies, the design of an accurate markerless system able to perform a quantitative clinical gait analysis is still an open challenge. Our purpose was to characterize the performance of a low-cost markerless system, consisting of the open-source library OpenPose, two webcams and a linear triangulation algorithm. The system was validated in terms of 3D kinematic gait analysis, by comparison with inertial sensors. Two synchronized videos of six healthy subjects were recorded in three webcam configurations, in walking and running sessions on a treadmill. Sagittal joint angles were compared between the two systems, to evaluate the kinematic performance of the markerless system. Results showed that the angular paths, provided by OpenPose, were close to the ones computed by inertial sensors in all camera positions, but OpenPose inaccuracy affected the computation of joint angles parameters producing absolute errors up to 14.0°± 1.8°. The observed differences depended on OpenPose ability in the recognition of joints centers and on camera configurations. In the optimal condition (joint visible from both cameras during the whole trial) our system accuracy resulted equal to 1.6 °± 0.3 °. This system could be useful in those applications where a real-time body tracking is required, i.e. video games or virtual rehabilitative trials. When a quantitative analysis is needed, the low level of accuracy may be overcome, by both increasing the cameras' number and implementing an optimization triangulation algorithm.

The use of motorized treadmills as convenient tools for the study of locomotion has been in vogue for many decades. However, despite the widespread presence of these devices in many scientific and clinical environments, a full consensus on their validity to faithfully substitute free overground locomotion is still missing. Specifically, little information is available on whether and how the neural control of movement is affected when humans walk and run on a treadmill as compared to overground. Here, we made use of linear and non-linear analysis tools to extract information from electromyographic recordings during walking and running overground, and on an instrumented treadmill. We extracted synergistic activation patterns from the muscles of the lower limb via non-negative matrix factorization. We then investigated how the motor modules (or time-invariant muscle weightings) were used in the two locomotion environments. Subsequently, we examined the timing of motor primitives (or time-dependent coefficients of muscle synergies) by calculating their duration, the time of main activation, and their Hurst exponent, a non-linear metric derived from fractal analysis. We found that motor modules were not influenced by the locomotion environment, while motor primitives were overall more regular in treadmill than in overground locomotion, with the main activity of the primitive for propulsion shifted earlier in time. Our results suggest that the spatial and sensory constraints imposed by the treadmill environment might have forced the central nervous system to adopt a different neural control strategy than that used for free overground locomotion, a data-driven indication that treadmills could induce perturbations to the neural control of locomotion.

Dynamic posturography combined with wearable sensors has high sensitivity in recognizing subclinical balance abnormalities in patients with Parkinson’s disease (PD). However, this approach is burdened by a high analytical load for motion analysis, potentially limiting a routine application in clinical practice. In this study, we used machine learning to distinguish PD patients from controls, as well as patients under and not under dopaminergic therapy (i.e., ON and OFF states), based on kinematic measures recorded during dynamic posturography through portable sensors. We compared 52 different classifiers derived from Decision Tree, K-Nearest Neighbor, Support Vector Machine and Artificial Neural Network with different kernel functions to automatically analyze reactive postural responses to yaw perturbations recorded through IMUs in 20 PD patients and 15 healthy subjects. To identify the most efficient machine learning algorithm, we applied three threshold-based selection criteria (i.e., accuracy, recall and precision) and one evaluation criterion (i.e., goodness index). Twenty-one out of 52 classifiers passed the three selection criteria based on a threshold of 80%. Among these, only nine classifiers were considered “optimum” in distinguishing PD patients from healthy subjects according to a goodness index ≤ 0.25. The Fine K-Nearest Neighbor was the best-performing algorithm in the automatic classification of PD patients and healthy subjects, irrespective of therapeutic condition. By contrast, none of the classifiers passed the three threshold-based selection criteria in the comparison of patients in ON and OFF states. Overall, machine learning is a suitable solution for the early identification of balance disorders in PD through the automatic analysis of kinematic data from dynamic posturography.