Gait symmetry methods: Comparison of waveform-based Methods and recommendation for use

Highlights

• All symmetry methods showed differences between the signals with different shapes.

• Cross-correlation method is sensitive to gait initiation phase.

• Cyclogram-based method is sensitive to signal scaling.

• Trend method and cross-correlation method are sensitive to signal phase shift.

Abstract

Gait symmetry has been shown to be a relevant measure for differentiating between normal and pathological gait. Although a number of symmetry methods exist, it is not clear which of these methods should be used as they have been developed using data collected from varying experimental protocols. This paper presents a comparison of state-of-the-art waveform-based symmetry methods and tests them on walking data collected from different environments. Acceleration signals collected from the ankle are used to analyse symmetry methods under different signal circumstances, such as phase shift, waveform shape difference, signal length (i.e. number of gait cycles) and gait initiation phase. The cyclogram based method is invariant to signal phase shifts, signal length and the gait initiation phase. The trend symmetry method is not affected by signal scaling and the gait initiation phase but is affected by signal length depending on the environment. Similar to the trend method, the cross-correlation symmetry method is not responsive to signal scaling and the gait initiation phase. The results of the symbolic method are not influenced by signal scaling, gait initiation and depending on the environment by the signal phase shift. From the results of the performed analysis, we recommend the trend method to gait symmetry assessment. The comparison of waveform-based symmetry methods brings new knowledge that will help in selecting an appropriate method for gait symmetry assessment under different experimental protocols.

Introduction

Bipedal gait is an essential human activity and is a complex movement that involves various interacting neuro-physiological systems [1]. Among a broad range of characteristics used to describe gait performance, gait symmetry has been shown to be a relevant measure in differentiating between healthy and pathological gait [2].

Data acquisition for gait symmetry assessment is often performed under different protocols [3], such as different walking speeds and distances [4], number of steps taken or gait cycles captured [5], etc. This inconsistency in protocols may not only influence symmetry results but also makes it difficult to evaluate which method should be used in a given situation. For example, healthy gait consists of three components: initiation, steady-state locomotion and termination. Although many studies have focused on steady-state gait [4,6], very few studies have evaluated gait initiation [7,8], and termination [9]. Most studies avoid them to exclude the effects of acceleration and deceleration, and focus only on the period of gait when the signal amplitudes vary the least. It was reported that gait initiation takes roughly between three to eight steps [10]. However, the first two steps are usually considered as accelerating and usually removed from analysis [11,12].

Quantitative gait symmetry methods can generally be divided into two categories: (1) methods that compare spatio-temporal parameters or symmetry indices, and (2) methods comparing entire gait cycles or waveform-based methods. While symmetry indices have proven very useful, waveform-based methods are potentially more informative [4,6]. Although many studies have compared different symmetry methods [4,6,13,14], they are restricted with respect to some gait pathology and consequently the symmetry methods are disease-specific.

Based on the above, this paper aims to compare well-known symmetry assessment methods under different circumstances. It focuses on demonstrating the performance of these methods on gait initiation phase, varying signal length (i.e. different number of gait cycles), phase shift, the signal’s shape and amplitude. As different walking conditions could potentially change walking behaviour [15], signals from different walking environments were used to examine the aforementioned symmetry methods. The comparison of waveform-based symmetry methods brings new knowledge that will help in selecting an appropriate method for gait symmetry assessment under different experimental protocols.