Gait characterization using wearable inertial sensors in healthy and pathological populations

Gait analysis is emerging as an effective tool to detect an incipient neurodegenerative disease or to monitor its progression. It has been shown that gait disturbances are an early indicator for cognitive impairments and can predict progression to neurodegenerative diseases. Furthermore, gait performance is a predictor of fall status, morbidity and mortality. Instrumented gait analysis provides quantitative measures to support the investigation of gait pathologies and the definition of targeted rehabilitation programs. In this framework, technologies such as inertial sensors are well accepted, and increasingly employed, as tools to characterize locomotion patterns and their variability in research settings. The general aim of this thesis is the evaluation, comparison and refinement of methods for gait characterization using magneto-inertial measurement units (MIMUs), in order to contribute to the migration of instrumented gait analysis from state of the art to state of the science (i.e.: from research towards its application in standard clinical practice). At first, methods for the estimation of spatio-temporal parameters during straight gait were investigated. Such parameters are in fact generally recognized as key metrics for an objective evaluation of gait and a quantitative assessment of clinical outcomes. Although several methods for their estimate have been proposed, few provided a thorough validation. Therefore an error analysis across different pathologies, multiple clinical centers and large sample size was conducted to further validate a previously presented method (TEADRIP). Results confirmed the applicability and robustness of the TEADRIP method. The combination of good performance, reliability and range of usage indicate that the TEADRIP method can be effectively adopted for gait spatio-temporal parameter estimation in the routine clinical practice. However, while traditionally gait analysis is applied to straight walking, several clinical motor tests include turns between straight gait segments. Furthermore, turning is used to evaluate subjects’ motor ability in more challenging circumstances. The second part of the research therefore headed towards the application of gait analysis on turning, both to segment it (i.e.: distinguish turns and straight walking bouts) and to specifically characterize it. Methods for turn identification based on a single MIMU attached to the trunk were implemented and their performance across pathological populations was evaluated. Focusing on Parkinson’s Disease (PD) subjects, turn characterization was also addressed in terms of onset and duration, using MIMUs positioned both on the trunk and on the ankles. Results showed that in PD population turn characterization with the sensors at the ankles lacks of precision, but that a single MIMU positioned on the low back is functional for turn identification. The development and validation of the methods considered in these works allowed for their application to clinical studies, in particular supporting the spatio-temporal parameters analysis in a PD treatment assessment and the investigation of turning characteristic in PD subjects with Freezing of Gait. In the first application, comparing the pre and post parameters it was possible to objectively determine the effectiveness of a rehabilitation treatment. In the second application, quantitative measures confirmed that in PD subjects with Freezing of Gait turning 360° in place is further compromised (and requires additional cognitive effort) compared to turning 180° while walking