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.

Parkinson’s disease (PD) is a disorder that causes a decrease in motor skills. Among the symptoms that have been observed, the most significant is the occurrence of Freezing of Gait (FoG), which manifests as an abrupt cessation of walking. This study investigates the impact of spatiotemporal gait parameters using wearable inertial measurement units (IMUs). Notably, 30 PD patients (15 with FoG, 15 without) and 20 healthy controls were enrolled. Gait data were acquired using two foot-mounted IMUs and key parameters such as stride time, gait phase distribution, cadence, stride length, speed, and foot clearance were extracted. Results indicated a tangible decline in motor abilities in PD patients, especially in those with FoG. Differences were observed in the segmentation of gait phases, with diminished swing phase duration observed in patients, and in the diminished spatial parameters of stride length, velocity, and foot clearance. Additionally, to validate the results, the accuracy of IMU-derived clearance measurements was validated against an optoelectronic system. While the IMUs accurately detected maximum points, the minimum clearance showed a higher measurement error. These findings support the use of wearable IMUs as a reliable and low-cost alternative to laboratory systems for the assessment of gait abnormalities in PD. Moreover, they highlight the potential for early detection and monitoring of FoG in both clinical and home settings.

A fully wearable solution estimating both kinematics and dynamics of motor tasks, including walking, running, and balance, would improve performance in clinical, rehabilitation, and sports settings. However, measuring GRF and CoP in a real-world setting remains an open challenge. Recently, sensorized insoles have become increasingly widespread, however, there are still many gaps in the determination of dynamic quantities. In this study, a novel composite insole with six FSRs was introduced to calculate GRF and CoP. The prototype was made with a commercial insole at the base, equipped with six FSRs and covered with two layers of silicone with different stiffness. The calibration curve was used to estimate GRF and CoP directly, and a machine learning approach was used to estimate them indirectly. As a first performance test, we conducted static standing trials. Due to the surface coverage of the sensors and the nature of the activity under study, the direct method showed underestimation of the GRF value. On the other hand, the ANN showed a superior performance in estimating GRF and CoP in subjects with regular foot but degrading in subjects with unusual foot. Preliminary results obtained on balance task are encouraging for future expansions of the training dataset, including other motor tasks.

Lower limb amputation causes a loss of motor and proprioceptive abilities in affected individuals. Different prostheses have been created to compensate for this deficit. However, most of these devices are passive and provide no sensory feedback to the amputee. Several solutions have been proposed to restore different typologies of sensation, driven by mechanical information obtained through sensors positioned on the anatomical district of interest, with specific limitations. This paper presents the mechanical and thermal characterization of a novel 3D-printed elastomeric strain sensor filled with liquid metal. It is designed to measure the strain on the cosmetic cover of a transtibial prosthesis to provide information to a feedback system. Static and dynamic tests were performed on the sensor, providing a graduation curve with satisfactory sensitivity, repeatability, linearity, and no signal attenuation over the 1 - 10 Hz frequency range. Thermal tests demonstrated an increase in the electrical resistance of the sensor up to 2.5% in the temperature range of 20 - 40 °C.

Training for confined-space operations presents significant challenges due to safety risks, limited accessibility, and the complexity of reproducing realistic conditions. Within this context, immersive technologies such as Head-Mounted Display Virtual Reality (HMD-VR) offer a promising alternative for procedural training. This study evaluates the effectiveness of a hands-free, bimanual HMD-VR training system for confined-space operations as an alternative to physical training when access to real environments is constrained. Two HMD-VR configurations were investigated and compared with a physical training condition using a shared multi-step procedural scenario. Training outcomes were assessed across repeated executions and through a real-world evaluation, focusing on procedural performance and cognitive demand. Results show that all groups achieved significant improvements in performance over repeated executions, with reductions in completion time and error rate. Importantly, no statistically significant differences were observed between groups during the final real-world evaluation, and non-inferiority analyses confirmed that HMD-VR training did not result in performance outcomes worse than those achieved through physical training within predefined operational margins. Similarly, no significant differences emerged in either EEG-derived Cognitive Load (CL) or subjective workload at the final execution, despite a reduction in perceived workload over time. Analysis of subjective immersive experience revealed consistently high levels of presence and engagement across HMD-VR conditions, with no significant differences between configurations. However, spatial presence and perceived realism decreased over repeated exposure, suggesting an adaptation effect rather than a continuous increase in immersion. Overall, the findings indicate that, under the controlled conditions of this study, HMD-VR training enables real-world procedural execution and cognitive demand profiles comparable to those observed following physical training. These results support the potential of HMD-VR as a viable approach for novice procedural training in confined-space operations, particularly in contexts where traditional training is limited by safety, cost, or accessibility constraints.