Use of non-linear metrics for the characterization of human motion: methodological constraints and functional interpretation

This PhD Thesis analysed the numerous and various metrics proposed for the quantification of motor stability in human motion analysis. Human motion analysis points to provide quantitative measures for the objective characterization of specific motion patterns, such as gait, aiming to support evidence based clinical decision. In recent years, the significant interest in finding effective methods for the quantification and prediction of fall risk in elderly subjects led to a proliferation of novel metrics. The majority of them originates from the theory of dynamical systems and has been used in robotics. Thus, they have been applied to gait analysis data, assuming similar interpretability in terms of motor control, resulting in a large amount of published studies, often leading to not conclusive and sometimes contrasting results. This can be related to the lack of a methodological reference for the appropriate experimental assessment and implementation of these metrics (e.g. target variables, number of strides, sampling frequency, implementation parameters) and of a clear functional correlate, establishing the relationship between the metrics and their possible clinical interpretation. Aiming to assess gait stability as an expression of motor control, both intrinsic properties of the human body and their relationship with the specific movement pattern must be taken into account. To this purpose, non-linear metrics were analysed (i.e. Lyapunov Exponent, Recurrence Quantification Analysis, Harmonic Ratio, and Multiscale Sample Entropy) describing different aspects of gait pattern related to the motor control system. The aim of this PhD dissertation was to improve the understanding of these non-linear metrics, providing evidence for the definition of methodological references for their experimental assessment, implementation, and possible clinical interpretation in specific conditions. Even though not exhaustive, the results provide an essential set of basic knowledge for the definition of a reference for the reliable use and interpretation of these non-linear metrics.