Validity and reliability of a portable gait analysis system for measuring spatiotemporal gait characteristics: comparison to an instrumented treadmill

Gait analysis serves as an important tool for clinicians and other health professionals to assess gait patterns related to functional limitations due to neurological or orthopedic conditions [1]. For instance, assessing and interpreting changes in spatiotemporal gait characteristics (e.g. gait speed, stride time, stride length, stride width, stride variability) has become important for predicting fall risk [2, 3]. These studies have been mainly conducted using stationary treadmills or portable optometric systems and gait mats [4]. However, this analysis is time-consuming and/or costly and obtained laboratory data may not be completely transferable to free-living conditions. For instance, elderly persons tend to walk at faster average speeds when walking longer (>20 m) than shorter distances (<10 m) [5] and parameters such as gait variability are more reliable when assessed for walking distances of at least 20 m [6, 7]. Hence, walking should be assessed outside the laboratory or on treadmills where persons can walk for extended periods of time. Moreover, simpler and low-cost mobile systems are desirable for evaluating pathological gait patterns outside the laboratory for treatment planning and evaluation.

Although body-worn sensor systems facilitate data collection outside the laboratory or clinic, most tests are usually performed in controlled settings. These laboratory or clinical environments for assessing gait are characterized by a level walkway, and patients are typically asked to walk at their habitual walking speed (normal speed). In contrast, free-living conditions include walking on varying slopes and surfaces and at different walking speeds resulting in locomotor adaptations including altered limb positioning and acceleration [8, 9]. To date, the effects of these changing environmental factors on the validity and reliability of body-worn sensor systems are not fully understood. Validity and reliability studies of gait analysis systems have to trade off the controlled setting of treadmill walking where gait parameters of many consecutive steps at predefined environmental conditions can be compared between systems and the real-time situation where gait parameters of only few isolated steps can be compared between different systems. Moreover, instrumented treadmills gait analysis is reliable [2] and considered a standard for evaluating the accuracy of portable and laboratory systems in measuring spatiotemporal gait characteristics in healthy subjects and patients [10‐12].

In recent years, activity monitors for classifying physical activity and pedometers for counting the number of steps have become available. Most accelerometer-based systems only provide step counts, stride times or cadence. To date, systems precisely capturing spatiotemporal gait characteristics are mostly used in specific (research) settings [13]. Novel and more complex systems not only comprise accelerometers but also gyroscopes and magnetometers, and hence are able to compute spatiotemporal and specific kinematic parameters. Such systems have been used to explore specific pathological gait and balance patterns [13‐15] and to evaluate training interventions [16, 17] in older adults and in patients with neurological diseases. RehaGait® (Hasomed, Magdeburg, Germany) is a commercial mobile system that specializes on measuring gait characteristics in field settings enabling gait data collection within less than 10 min (preparation: 1 min; calibration: 1 min; data collection: 1–8 min. The system comprises multi-directional accelerometers, gyroscopes and magnetometers, software computing spatiotemporal gait parameters as well as selected joint kinematics during overground or treadmill walking and running, a large database of normative gait data (more than 2000 healthy subjects ranging in age from 5 to 90 years) and an intuitive and comprehensive tablet user interface representing a promising tool for therapists. However, to date only limited data on the validity and reliability of RehaGait® are available [18]. Specifically, to date the validity of the RehaGait® for measuring gait characteristics of treadmill walking has not been determined. This information is critical for its potential application in research and clinical practice. Besides the parameters investigated here, the system also computes gait phases, foot to ground angle, foot height and hip circumduction.

The purpose of this study was to assess the validity of the body-worn sensor system RehaGait® for measuring spatiotemporal gait characteristics with a stationary treadmill and the reliability of both systems at different speeds and slopes.

The purpose of this study was to assess the validity of the RehaGait® with a stationary treadmill and the reliability of both systems at different velocities and slopes. We found good to excellent validity for stride length, cadence and stride time between the RehaGait® system and the instrumented treadmill in healthy younger adults. The RehaGait® system overestimated stride length (+2.7 %) and stride time (+0.8 %) and underestimate cadence (−1.5 %) with small to moderate effect sizes for all speeds and slopes. ICCs were slightly lower at slow compared to normal and fast walking speeds which is in agreement with previous findings [12]. Low day-to-day variability indicates good to excellent reliability of the RehaGait® system. Larger limits of agreement for walking at 15 % slope suggests that uphill walking may influence the reliability of the RehaGait® system.

Although significant differences in all gait characteristics were found between the RehaGait® and the instrumented treadmill, these differences were smaller than differences in these gait characteristics between other body-worn gyroscope based sensors and the GAITRite® system [28]. In our study, the average difference in stride length was less than 5 cm for walking at normal speed at 0 % slope compared to almost 8 cm reported by Greene et al. [28]. The latter study assessed spatiotemporal gait characteristics at a much greater range of walking speeds (0.89–1.72 m/s) than our study (0.91–1.30 m/s), and the normal speed in the study by Greene et al. corresponds to the fast speed in our study. While different environmental factors affect gait, walking speed is mainly modulated by altering stride length and only small changes in cadence [29]. Differences in spatiotemporal gait characteristics between accelerometer based gait analysis and the GAITRite® of less than 0.02 m/s walking speed, 1 cm step length and 2 ms step time in older adults have been reported [30]. The agreement between data obtained with the RehaGait and the instrumented treadmill was better by a factor of 10 for stride time and worse by a factor of 2 than that of other wearable technology [12]. Hence, the RehaGait® can be used to assess spatiotemporal gait characteristics of level treadmill walking with sufficient accuracy, although the agreement of temporal parameters is better than that of spatial and spatiotemporal parameters shown by the RMS errors. However, we cannot elucidate conclusively if the greater agreement between data measured using RehaGait® and the instrumented treadmill compared to that of other portable systems can be attributed to improved algorithms or technology or to differences in methodology.

To date, only few studies have investigated the effect of inclined walking at different speeds on basic spatiotemporal gait characteristics. For instance, Leroux et al. [23] reported that postural adaptations to increasing walking surface slope are accompanied by gradual increases in stride length as the uphill slope becomes steeper. In our study, we also observed greater stride length when walking on a 15 % slope compared to level walking. Interestingly, the increases in stride length detected by the RehaGait® were greater than those detected by the instrumented treadmill resulting in lower validity for walking on 15 % slope than on 0 % slope for the RehaGait®. One possible explanation for this discrepancy is that – albeit not recorded in this study – the foot striking pattern may change when walking uphill potentially affecting the accuracy of identifying gait events by both systems. For instance, foot strike may transition from foot strike to midfoot strike when walking uphill hence potentially altering acceleration (inertial sensor) and pressure (instrumented treadmill) patterns. Hence, such changes in gait mechanics may affect the accuracy of identifying foot strike for both systems hence potentially affecting spatial and spatiotemporal gait parameters. Our results confirm previously reported [31] decreases in cadence with increasing uphill slope only for slow speeds. However, in our study speed was kept constant for walking at both slopes, and hence participants were not able to freely adjust their gait patterns to the changing environment.

Other systems comprising inertial sensors mounted to the foot have been shown to provide accurate estimates of walking speed and incline for walking at a range of speeds and inclines, respectively [32]. The purpose of our study was to test the system validity of the RehaGait® system, and hence we only tested two inclines commonly used in clinical environments [20, 21]. The fact that walking speed recorded by the RehaGait® adequately corresponded to the speed set on the treadmill tachometer at flat and inclined slope emphasizes the validity of this portable gait analysis system for measuring walking speed for treadmill walking. Moreover, the large discrepancy in walking speed set on the treadmill tachometer and that indirectly calculated from the built in pressure mat suggests that the tachometer speed of the instrumented treadmill is more reliable than the calculated speed.

Several factors may have contributed to the differences in gait characteristics between RehaGait® and the instrumented treadmill. First, the two systems measure two different quantities. The RehaGait® measures the acceleration and angular velocity of the foot while the instrumented treadmill measures the pressure distribution under the foot. Based on these different quantities, specific algorithms are employed to calculate gait events that may be differently affected by factors such as foot placement. Slight differences in these definitions may cause systematic differences in gait characteristics between the two systems. The double integration of the acceleration signal may affect the calculation of spatiotemporal gait characteristics such as walking speed and stride length and hence explain the slightly lower validity for these characteristics. However, because potential drift is offset during zero acceleration phases of the sensor, this effect is expected to be minimal. Moreover, stride length assessed by the instrumented treadmill was calculated as the distance between two initial heel pressure points of alternate sides. Hence, even small step-to-step variability in foot placement (heel-strike versus midfoot-strike) would contribute to inaccuracies in calculated stride length and hence also in calculated walking speed explaining the discrepancy between latter and the tachometer speed.

In this study, we compared gait characteristics between the RehaGait® and an instrumented treadmill. The advantage of the RehaGait® over laboratory based 3days gait analysis systems is that the RehaGait® can be used to assess gait in free-living conditions. While treadmill walking differs from overground walking [33], a previous study [6] has shown that body worn sensor technology provide valid and reliable gait characteristics only for gait that is performed over walking distances exceeding 20 m. In our study, we measured gait data for approximately 200 strides, and a validation for gait characteristics for such distances can only be performed using a treadmill. Moreover, the use of a treadmill for validating the RehaGait® enabled us to control the walking speeds both for level walking and for walking on an incline. In this study, we only included healthy younger subjects and hence cannot make a statement regarding validity of the RehaGait® in other populations such as older adults or people with pathologies.

Overall, our results further support the use of inertial sensor based gait analysis system with associated advantages of facilitating cost and time efficient assessment of gait patterns. However, the applicability of such systems depends on the clinical and research question because only selected gait parameters can be assessed. For instance, current inertial sensor based systems cannot measure stride width, a parameter that has been identified as being important for predicting fall risk [2, 3]. Future research is warranted to elucidate other surrogate measures for risk of falls and of neuromuscular and musculoskeletal conditions that can be assessed using these novel sensors.

RehaGait® is a commercial mobile system that specializes on measuring gait characteristics in field settings enabling gait data collection within less than 10 min. We showed good to excellent validity for walking speed, stride length, cadence and stride time between the RehaGait® system and the instrumented treadmill in a convenience sample of healthy younger adults with good to excellent reliability. Larger limits of agreement for walking at 15 % slope than at 0 % slope suggests that walking slope may influence the reliability of the RehaGait® system. Spatiotemporal gait information obtained with this system can be used for functional evaluations of patient populations augmenting clinical assessments for treatment planning and evaluation.