Reliability of segmental accelerations measured using a new wireless gait analysis system

Abstract

The purpose of this study was to determine the inter- and intra-examiner reliability, and stride-to-stride reliability, of an accelerometer-based gait analysis system which measured 3D accelerations of the upper and lower body during self-selected slow, preferred and fast walking speeds. Eight subjects attended two testing sessions in which accelerometers were attached to the head, neck, lower trunk, and right shank. In the initial testing session, two different examiners attached the accelerometers and performed the same testing procedures. A single examiner repeated the procedure in a subsequent testing session. All data were collected using a new wireless gait analysis system, which features near real-time data transmission via a Bluetooth network. Reliability for each testing condition (4 locations, 3 directions, 3 speeds) was quantified using a waveform similarity statistic known as the coefficient of multiple determination (CMD). CMD's ranged from 0.60 to 0.98 across all test conditions and were not significantly different for inter-examiner (0.86), intra-examiner (0.87), and stride-to-stride reliability (0.86). The highest repeatability for the effect of location, direction and walking speed were for the shank segment (0.94), the vertical direction (0.91) and the fast walking speed (0.91), respectively. Overall, these results indicate that a high degree of waveform repeatability was obtained using a new gait system under test-retest conditions involving single and dual examiners. Furthermore, differences in acceleration waveform repeatability associated with the reapplication of accelerometers were small in relation to normal motor variability.

Introduction

Relative to other motion measurement devices, accelerometer-based gait analysis systems have several advantages. Accelerometer-based systems are typically lightweight and portable, which facilitate unencumbered movement of the subject and do not confine data collection to the laboratory environment. Additionally, they are easy to use, cost effective, are able to capture data from many gait cycles, and avoid errors associated with differentiation of raw displacement data. Importantly, accelerometers are sufficiently sensitive to detect rapid movements even when the corresponding displacements are small, such as during physiological or pathological tremor (Newell and Sprague, 1996; Vaillancourt and Newell, 2000), or the rapid lateral motion of an unstable knee (Yoshimura et al., 2002). Together these advantages have meant that accelerometers have been useful in a wide range of gait-related settings. Previous applications include quantification of physical activity levels (Aminian et al., 1999; Hoos et al., 2004; Zhang et al., 2003), determining spatial-temporal gait variables (Mansfield and Lyons, 2003; Moe-Nilssen and Helbostad, 2004; Zijlstra and Hof, 2003), estimation of hip joint loading patterns (van den Bogert et al., 1996; Zijlstra and Bisseling, 2004), and the assessment of balance and stability during locomotion (Menz et al., 2003a, Menz et al., 2003b; Moe-Nilssen, 1998a). These applications are of particular use in clinical and rehabilitation settings, as they provide a simple means of obtaining quantitative gait data that can be used for diagnostic purposes and to monitor treatment progress.

In spite of the many advantages of using accelerometers in gait studies, there are a number of limitations associated with their use. For example, it is not straightforward to distinguish between the inertial and gravitational components of the signal without additional information describing the orientation of the device. There are also difficulties relating acceleration data to a global reference frame since the acquired data's frame of reference is continually moving. Additionally, skin motion artifact due to impact loading and muscle activation can readily contaminate signals. The anatomical location and orientation of the attached accelerometers can also have a substantial influence on the output signal. This issue is particularly relevant if the examiner attaching the devices has limited experience or if different examiners attach the devices under retest conditions. Together these limitations have the potential to compromise the quality of acceleration data recorded from body-mounted sensors during gait, especially under test–retest conditions. Therefore, efforts to determine the reliability of gait data obtained from accelerometer-based systems are warranted.

Few studies have assessed the reliability of accelerations recorded using body-mounted sensors. Moe-Nilssen (1998b) and Henriksen et al. (2004) reported good test–retest reliability for accelerations measured during walking using a single triaxial accelerometer attached to the lower trunk. In these investigations, root-mean square (RMS) accelerations were calculated for walking trials conducted over separate days, and the reliability computed using the intraclass correlation coefficient (ICC). ICC values for the mean accelerations were reported to range from 0.77 to 0.96 for the three orthogonal axis directions. Similar results were briefly reported by Menz et al. (2003a) for patterns of head and trunk RMS accelerations during level walking, with ICC's ranging from 0.84 to 0.97. From these studies it is apparent that RMS accelerations calculated from accelerometry data during walking are reliable across testing sessions. However, it remains unclear whether differences exist in reliability across testing sessions when accelerations are measured from different locations on the body. This is important for investigations, which examine relationships between accelerations measured from multiple body segment locations. Examples of such studies include those that focus on patterns of coordination between the head and trunk during walking (Kavanagh et al., 2004, Kavanagh et al., 2005), and/or those concerned with quantifying attenuation of gait-related accelerations between the lower and upper body (Hamill et al., 1995; Voloshin and Wosk, 1982). Arguably, a more comprehensive approach for examining reliability of 3D accelerations during gait is by using a waveform reliability statistic that takes into account the time-series evolution of the signal, instead of the mean signal amplitude computed over several gait cycles. One such statistic, known as the coefficient of multiple determination (CMD), has been highlighted as a useful measure of determining reliability of kinematic and kinetic gait data obtained under test–retest conditions (Besier et al., 2003; Kadaba et al., 1989).

The purpose of this study was to determine inter- and intra-examiner reliability, and stride-to-stride reliability, of acceleration data collected from multiple locations on the body at a range of walking speeds. It was hypothesized that stride-to-stride reliability would be higher than inter- or intra-examiner reliability due to the errors associated with the reapplication of accelerometers. In addition, we describe and outline the advantages of a new accelerometer-based gait system that features real-time data transmission via a Bluetooth network protocol.