Inertial measurement units (IMUs) have a long-lasting popularity in a variety of industrial applications from navigation systems to guidance and robotics. Their use in clinical practice is now becoming more common, thanks to miniaturization and the ability to integrate on-board computational and decision-support features. IMU-based gait analysis is a paradigm of this evolving process, and in this study its use for the assessment of Parkinson's disease (PD) is comprehensively analyzed. Data coming from 25 individuals with different levels of PD symptoms severity and an equal number of age-matched healthy individuals were included into a set of 6 different machine learning (ML) techniques, processing 18 different configurations of gait parameters taken from 8 IMU sensors. Classification accuracy was calculated for each configuration and ML technique, adding two meta-classifiers based on the results obtained from all individual techniques through majority of voting, with two different weighting schemes. Average classification accuracy ranged between 63% and 80% among classifiers and increased up to 96% for one meta-classifier configuration. Configurations based on a statistical preselection process showed the highest average classification accuracy. When reducing the number of sensors, features based on the joint range of motion were more accurate than those based on spatio-temporal parameters. In particular, best results were obtained with the knee range of motion, calculated with four IMUs, placed bilaterally. The obtained findings provide data-driven evidence on which combination of sensor configurations and classification methods to be used during IMU-based gait analysis to grade the severity level of PD.
The widespread and pervasive use of smartphones for sending messages, calling, and entertainment purposes, mainly among young adults, is often accompanied by the concurrent execution of other tasks. Recent studies have analyzed how texting, reading or calling while walking–in some specific conditions–might significantly influence gait parameters. The aim of this study is to examine the effect of different smartphone activities on walking, evaluating the variations of several gait parameters. 10 young healthy students (all smartphone proficient users) were instructed to text chat (with two different levels of cognitive load), call, surf on a social network or play with a math game while walking in a real-life outdoor setting. Each of these activities is characterized by a different cognitive load. Using an inertial measurement unit on the lower trunk, spatio-temporal gait parameters, together with regularity, symmetry and smoothness parameters, were extracted and grouped for comparison among normal walking and different dual task demands. An overall significant effect of task type on the aforementioned parameters group was observed. The alterations in gait parameters vary as a function of cognitive effort. In particular, stride frequency, step length and gait speed show a decrement, while step time increases as a function of cognitive effort. Smoothness, regularity and symmetry parameters are significantly altered for specific dual task conditions, mainly along the mediolateral direction. These results may lead to a better understanding of the possible risks related to walking and concurrent smartphone use.
The accurate and reliable extraction of specific gait events from a single inertial sensor at waist level has been shown to be challenging. Among several techniques, a wavelet-based method for initial contact (IC) and final contact (FC) estimation was shown to be the most accurate in healthy subjects. In this study, we evaluated the sensitivity of events detection to the wavelet scale of the algorithm, when walking at different speeds, in order to optimize its selection. A single inertial sensor recorded the lumbar vertical acceleration of 20 subjects walking at three different self-selected speeds (slow, normal, and fast) in a motion analysis lab. The scale of the wavelet method was varied. ICs were generally accurately detected in a wide range of wavelet scales under all the walking speeds. FCs detection proved highly sensitive to scale choice. Different gait speeds required the selection of a different scale for accurate detection and timing, with the optimal scale being strongly correlated with subjects’ step frequency. The best speed-dependent scales of the algorithm led to highly accurate timing in the detection of IC (RMSE < 22 ms) and FC (RMSE < 25 ms) across all speeds. Our results pave the way for the optimal adaptive selection of scales in future applications using this algorithm.