Validation of a system to measure gait cycle parameters by wearable sensors and computational methods

Abstract

Gait abnormalities in Normal Pressure Hydrocephalus (NPH) and Parkinson’s Disease (PD) patients share similar characteristics. However, certain quantities differ between the two diseases. In particular, gait asymmetry, due to the difference between the movement of the left and right limbs during walking, is known to be more common in PD. To compare gait between PD and NPH, within the MoveSenseAI (MSAI) study, of which this thesis is part, we use prototype wearable sensors (ActiSense, IEE) to measure time-series data which we then employ to compute gait cycle parameters such as cadence, gait cycle duration, stance phase and swing phase. This system is the foundation of the MSAI study which has the final objective of developing a Machine Learning (ML) algorithm to classify the disease according to the patients’ gait. The purpose of this thesis is twofold: the validation of a) the system employed within MSAI, composed of the prototype sensors and a newly developed computational pipeline, and the validation of b) its capability to detect gait asymmetry. For this purpose, 2 different groups of respectively 9 and 12 healthy subjects were recruited, in order to measure their gait and a) compare it to the measurements performed by means of an instrumented treadmill (Gaitway 3D, h/p/cosmos) representing the golden standard in the field and b) analyze gait asymmetry using standardized clinical tests. For objective a) 9 healthy subjects, while equipped with ActiSense, performed 3 walks on Gaitway 3D at the speeds of 2, 3 and 4 km/h, each for a duration of 60 seconds with both systems recording in parallel. For objective b) we measured the gait of 12 healthy subjects considered as controls and of the same subjects performing the walking tests with loads added, alternatively, to one or the other of their lower limbs, to artificially induce asymmetry in their gait in a controlled and reproducible manner. The results from part a) gave us insight into how to calibrate the estimation of the gait cycle parameters by comparing the results of both systems with statistical analysis, ultimately allowing us to validate our system. The results of part b) assessed the potential of the system to detect asymmetry based on symmetry ratios of gait cycle parameters which will subsequently be applied to data from PD and NPH patients within the MSAI study.