Phd in INDUSTRIAL AND MANAGEMENT ENGINEERING

Thesis title: Wearable sensors for human kinematics and kinetics assessment: Ergonomics, Clinical and Rehabilitation applications

Wearable devices are the most promising solution for evaluating injuries and disabilities due to occupational hazards, and for the assessment of motor impairment due to neuro-muscular diseases. The main advantage of this new technology allows to overcome the limits of the diagnosis assessment, which is traditionally conducted in clinical setting, and to perform motion analysis outdoor reducing time-consuming processes related to the traditional instrumentation. However, wearable sensors, such as inertial sensors, need to be validated to obtain accurate and reliable measurements of the human kinematics and kinetics. Moreover, they need to be combined with reliable biomechanical models, which reproduce human kinematics, and machine learning algorithms, to achieve the purpose of the real-time monitoring. In this thesis, the application of wearable sensors was explored in ergonomics, to assess the influence of awkward postural patterns, to highlight the effectiveness about the integration of a full setup in workplace for the prevention of work-related musculoskeletal disorders with respect to a small setup. They were also used to assess human dynamics during manual material handling tasks, in combination with another wearable solution, i.e., pressure insoles. In clinical and rehabilitation field, instead, the wearable sensors represent a valid substitution of the traditional devices and an innovation in the assessment of motor impairment and rehabilitation follow-up. Measurements of the kinematic or balance parameters allow to verify the benefit of the therapies and the recovery time related with the motor deficit, guaranteeing the assistance of the patient during the therapies remotely. Furthermore, as per the ergonomics field, a wearable solution could help to obtain measurements of human motion to improve or avoid the incorrect estimation based on clinical scales. This thesis work firstly explores the potential of wearable sensors combined with a biomechanical model, which reconstruct the human kinematics, to discriminate two postural patterns, i.e., squat and stoop lifting strategy, during manual material handling tasks through the evaluation of kinematic parameters of the upper body and lower limbs. The statistical results unveil the difference between correct and incorrect postural patterns for all estimated kinematic parameters (p<0.01), and changes related to the weight of the load for the hips joint and trunk kinematics (p<0.01). Subsequently, it combines the data obtained with a machine learning algorithm, the support vector machine, to demonstrate that the kinematics parameters estimated allow to automatically distinguish an awkward posture from a safety postural pattern and to highlight the inefficacy to reduce the number of sensors to quantify a biomechanical load during lifting load tasks. In fact, an accuracy of 99.4% was achieved with a dataset, that refers to the kinematics obtained through a full setup, in contrast with the low value of accuracy (76.9%) obtained through the small setup, that refers only to the kinematics of the trunk body segment. As a third step of this ergonomics study, the attention was focused on the assessment of human dynamics, combining two wearable systems, i.e., the inertial sensors and the wearable pressure insoles, to apply the bottom-up inverse dynamics method. Considering the gaps and different methods in literature, the two systems were validated to quantify the accuracy, repeatability, and reliability of the estimation of the forces exerted at level of the joints through the wearable solution, especially the L5/S1 joint, where the overexertion and stress on the back are responsible of the low back pain. The results obtained for the anteroposterior and vertical components of the forces of all the lower limb joints and L5/S1 joint are encouraging towards the adoption of the wearable solution. On the contrary, the lack of measurement of the mediolateral component of the GRFs through the wearable pressure insoles, does not allow to estimate such a component in the joint forces. Regarding the clinical and rehabilitation application, the wearable sensors were adopted to assess the effect of patellar taping on kinematics of lower limb joints, in particular the knee joint, and the human Center of Mass (CoM). In the study related to the taping effect on kinematics, the reduction of the range of motion of the knee joint due to the taping allow to quantify through the wearable system the benefit of the taping for the Single Leg Squat task (p=0.01), while for the Star Excursion Balance Test a similar trend was observed. Considering the second application of the wearable sensors in this part of the study, two methods based on wearable sensors were validated through the gold standard in terms of accuracy e reliability of the measurements of the human CoM displacement in balance and dynamic tasks. Moreover, the effectiveness of the combination of biomechanical model and a full wearable setup, in comparison to a Strap Down Integration method with only one sensor, was investigated. High correlation coefficients were obtained between the CoM displacement estimated through the biomechanical model and the gold standard, while the Strap Down Integration method showed low values of accuracy.