Automated IMU-based Motion Analysis for Clinical Applications: the Parkinson's Disease and Post-stroke Cases

Abstract

Motion is a fundamental component of our everyday life. Population ageing, in place especially in the developed countries in the last decades, has made increasingly present the issues related to elderly people's mobility as preserving the ability to move without difficulties and independently is crucial for maintaining a good life quality level. Therefore, the interest of the scientific community in motor-impairing diseases which are more likely to occur in the elderly population, such as Parkinson's Disease (PD) and stroke, has grown steadily in recent years.

The analysis of motor performance is a key aspect of many clinical assessment processes but often clinicians can rely only of qualitative analysis and on their experience. Technology can contribute consistently in helping medical personnel to provide more accurate and reliable diagnoses and to define more effective therapies. Human motion analysis techniques, for example, are often used in clinical applications to quantitatively characterize patients' movements during specific motor tasks. Classical motion analysis systems, based on optical, magnetic, or mechanic technologies, are accurate and reliable but, on the other hand, they are expensive, require external infrastructures and specifically trained personnel. Inertial Measurement Units (IMUs)-based motion analysis systems are lately becoming the most adopted alternative because they are easy-to-use, flexible, cost-effective, unobtrusive and can be used in free living environment.

In this thesis, a unified approach, relying on IMU-based motion sensing, has been used to develop a system for the automatic assessment of functional motor tasks, with particular focus on gait, in both PD and post-stroke populations. The proposed systems have been designed to be: (i) low-cost and low complexity, using Body Sensor Networks (BSNs) formed by only few inertial nodes; (ii) accurate and objective in the kinematic characterization of specific motor tasks (validation of the developed algorithms through comparison with gold standard technologies); (iii) "smart", exploiting machine learning techniques for providing motor performance automatic scoring, which mimic clinicians' evaluation criteria, together with meaningful high level and aggregate information; (iv) tele-health ready, being suitable for remote and long-term monitoring of patients outside the clinics. The results obtained in both systems demonstrate that IMU-based motion analysis, especially when supported by machine learning techniques, represents a powerful tool for enhancing the quality of the clinical assessment process both within and outside the clinics.