Agreement between BTS G-walk and GaitLab in spatiotemporal and pelvic angle measurements in active older adults

Introduction

Aging is a natural process that affects all individuals, leading to physiological changes that impact health and well-being, particularly in older adults. One of the most significant alterations occurs in gait patterns, which reflect mobility, functional independence, and overall health status in this population (Cruz-Jimenez, 2017). Age-related declines, such as reduced walking speed, shorter step length, and increased double support time, are associated with a higher risk of falls, disability, and mortality (Dewolf et al., 2021; Hollman, McDade & Petersen, 2011), underscoring the importance of gait assessment for identifying abnormalities, understanding the effects of aging on locomotion, and monitoring interventions aimed at improving functionality in older adults (Verghese et al., 2006; Aboutorabi et al., 2016). In this context, regular physical activity plays a crucial role in positively influencing gait patterns, especially in physically active older adults, whose gait characteristics may differ due to enhanced physical conditioning (Sethi, Bharti & Prakash, 2022).

According to the World Health Organization (WHO), an older adult is considered physically active if they engage in at least 150 min of moderate-intensity aerobic activity, 75 min of vigorous-intensity aerobic activity, or an equivalent combination per week, in addition to two strength training sessions targeting major muscle groups (Rudnicka et al., 2020). This level of activity has been linked to attenuated declines in gait, such as reduced speed loss and greater stability (Boyer, Andriacchi & Beaupre, 2012; Paterson & Warburton, 2010). However, normative data and specific validation studies for this group remain scarce, limiting the understanding of gait patterns in active older adults and highlighting the need for precise assessment tools (Mehdizadeh et al., 2022). In this regard, three-dimensional (3D) motion capture systems are widely recognized as the gold standard for gait analysis, providing detailed data on spatiotemporal, kinematic, and kinetic parameters, including joint angles and ground reaction forces (Kadaba et al., 1989; Jakob et al., 2021). These systems use high-frequency infrared cameras and reflective markers, often following the Helen Hayes protocol (Kadaba, Ramakrishnan & Wootten, 1990; Simon, 2004).

The classification of three-dimensional (3D) motion capture systems as the gold standard for gait analysis is supported by a consistent body of evidence across different methodological designs. Donno et al. (2023) conducted a multifactorial characterization of gait in young adults using an optoelectronic BTS SMART-DX system coupled with force plates, highlighting its capacity to capture fine-grained kinematic and kinetic data, including joint moments and powers during both forward and backward walking. The authors emphasized the clinical value of such systems, especially when detailed biomechanical insight is required for rehabilitation planning or pathological gait assessment. Similarly, Lee et al. (2013) used a Vicon 3D motion capture setup with synchronized force platforms to compare forward and backward locomotion, demonstrating the system’s ability to quantify critical joint behavior and power dynamics with high precision across multiple plane. Furthermore, Bovi et al. (2011) presented comprehensive normative kinematic, kinetic, and electromyographic datasets for multiple locomotor tasks in both young and older adults using a SMART-E system. Their study reinforced the reliability of 3D motion capture for constructing reference models and detecting subtle gait deviations under different functional conditions. Together, these findings support the use of 3D motion capture as a methodological benchmark for validating emerging gait analysis tools.

Despite their robustness, the complexity and high costs of 3D motion capture systems make them impractical for large-scale clinical use, particularly outside specialized laboratories (Scataglini et al., 2024). This has driven the development of alternative technologies, such as wearable inertial sensors like the BTS G-Walk, which offer greater accessibility and portability (Fernández-Gorgojo et al., 2022). Equipped with accelerometers, gyroscopes, and magnetometers, these devices measure spatiotemporal parameters and pelvic angles in real time, making them applicable in both controlled and natural settings (Tao et al., 2012; Iosa et al., 2016). While some studies have compared inertial sensors with other gait analysis devices, few have investigated their agreement with 3D motion capture systems, and none have specifically assessed hip angulations in physically active older adults (Yazıcı, Co˛banoğlu & Yazıcı, 2022; Salarian et al., 2004). This gap limits the validation of these devices for populations with distinct gait characteristics resulting from regular physical activity (Jarchi et al., 2018; Cappozzo et al., 1996). Older adults represent a clinically relevant population for gait assessment, as walking speed and step parameters are strongly associated with functional independence, frailty, and fall risk (Studenski et al., 2011). Accessible tools that allow frequent and reliable monitoring of these parameters are particularly valuable in this group (Fritz & Lusardi, 2009). Therefore, this study aimed to validate the BTS G-Walk system specifically in physically active older adults, who stand to benefit most from simplified gait assessments that support functional health strategies.

Thus, this study aimed to evaluate the agreement between the BTS G-Walk inertial sensor and the BTS GaitLab motion capture system in measuring spatiotemporal parameters and pelvic angles during gait in physically active older adults. The goal was to provide evidence supporting the validity of the G-Walk as a practical and reliable tool for gait analysis, with potential applications in mobility monitoring and intervention planning in gerontology.

Material & methods

Results

Table 2 presents the agreement between the BTS GaitLab (Method 1) and the BTS G-Walk (Method 2) across spatiotemporal and pelvic parameters. The spatiotemporal parameters showed strong agreement between the two systems. Cadence and walking speed demonstrated excellent intraclass correlation coefficients (ICC = 0.98 and 0.96, respectively), with no statistically significant differences between the methods (p = 0.215 and p = 0.207), and negligible effect sizes (Cohen’s d = 0.16 and 0.17), suggesting high consistency and minimal bias. Step length and stride length also presented high agreement (ICC ranging from 0.87 to 0.92). Although both parameters showed statistically significant differences (p < 0.05), effect sizes were small to moderate (d = 0.28–0.49), indicating that these discrepancies, while statistically detectable, are unlikely to be clinically meaningful.

In contrast, temporal parameters exhibited lower agreement. Stance and swing times showed ICCs ranging from −0.54 to 0.26, indicating weak or absent agreement between the systems. These variables also presented statistically significant differences (p < 0.001), with moderate to large effect sizes (d = 0.62 to 0.84), reflecting a consistent divergence in measurements. Double support time had the lowest agreement among temporal variables, with a large effect size (d = 0.91), further highlighting discrepancies in the temporal domain between the two systems.

Pelvic kinematic variables demonstrated variable agreement depending on the axis of measurement. Pelvic tilt and obliquity showed moderate ICCs (0.48 to 0.78), with statistically significant differences (p < 0.01) and large effect sizes (Cohen’s d ranging from 0.74 to 1.09), suggesting that although the direction of movement was similarly captured, the magnitude differed substantially between systems. Pelvic rotation presented the lowest ICCs (0.28 to 0.38), with no statistically significant differences between methods (p = 0.284 and p = 0.232) and small effect sizes (d = 0.14 and 0.16), suggesting random variation and poor agreement in this axis.

Bland–Altman analyses visually supported the numerical findings. For spatial parameters such as cadence, speed, and step length, the plots (Figs. 1 to 6) revealed narrow limits of agreement, minimal bias, and a low proportion of outliers (0% to 5%), indicating strong agreement and low dispersion between systems. In contrast, the plots for temporal parameters (Figs. 7 to 14), including stance, swing, and particularly double support time, showed broader limits of agreement and more scattered data—mirroring the lower ICC values and higher effect sizes observed, with double support time displaying the greatest discrepancy. Similarly, pelvic angle plots (Figs. 15 to 20) highlighted the variability across axes: tilt and obliquity showed systematic bias but moderate agreement, whereas pelvic rotation exhibited high dispersion and poor consistency between systems.

Table 3 presents intra-method variability, assessed by the coefficient of variation (CV). Temporal parameters exhibited the highest variability, especially double support time, where CV values ranged from 20.1% to 20.9%.

Discussion

The findings of this study revealed that the BTS G-Walk inertial sensor exhibits high agreement with the BTS GaitLab motion capture system in measuring spatial gait parameters in physically active older adults, including cadence (ICC = 0.98), walking speed (ICC = 0.96), and step/stride length (ICC = 0.90–0.92), across a robust sample of 59 participants. These measures showed small, non-significant mean differences (p > 0.05 for cadence and speed), and small effect sizes (Cohen’s d = 0.16 and 0.17), indicating that the G-Walk provides reliable estimates comparable to a gold-standard 3D system for spatial gait characteristics. This consistency aligns with previous studies, such as Vítečková et al. (2020), who reported high agreement between the G-Walk and another validated system in healthy adults and Parkinson’s patients, and De Ridder et al. (2019), who observed similar precision in young individuals. The robustness of these variables may stem from the ability of accelerometers and gyroscopes to capture repetitive movement patterns with less reliance on fine temporal resolution, as noted by Tao et al. (2012) in a comprehensive review of wearable sensors. These results support the use of the G-Walk in clinical and research settings where laboratory-based systems are unavailable, particularly for monitoring functionality in active older adults, a population in which speed and cadence are established indicators of locomotor health (Hollman, McDade & Petersen, 2011). The narrow limits of agreement and minimal bias observed in the Bland–Altman plots further reinforce the consistency of spatial parameters.

In contrast, temporal gait parameters—such as stance time (ICC = −0.18 to 0.26), swing time (ICC = −0.54 to 0.17), and single/double support times (ICC = −0.11 to 0.32) demonstrated weak to no agreement, with statistically significant differences (p < 0.001 in most cases). These ICC values, some even negative, suggest not only measurement variability but also a fundamental inconsistency in gait event detection between the systems. A likely explanation lies in the differing sampling frequencies (100 Hz for the G-Walk versus 200 Hz for the GaitLab cameras and 400 Hz for force plates), which may result in less precise identification of key events like heel strike and toe-off. Additionally, the proprietary gait event detection algorithms of each system may employ distinct criteria for defining gait phases, further contributing to discrepancies, as suggested by Jakob et al. (2021) in a validation with Parkinson’s patients. Intra-observer variability (Table 3) reinforces this limitation, with higher coefficients of variation for double support time (15.3% for GaitLab and 20.1% for G-Walk), a critical stability indicator in older adults (Verghese et al., 2006). The wider dispersion observed in the Bland–Altman plots, especially for double support time, visually confirms the numerical inconsistencies in temporal parameters. These findings highlight a limitation of the G-Walk for detailed temporal assessments, particularly in populations where gait alterations are clinically significant.

The analysis of pelvic angles, a novel aspect of this study, showed moderate agreement for tilt and obliquity (ICC = 0.48 to 0.78), with statistically significant differences (p < 0.001) and large effect sizes (Cohen’s d ranging from 0.74 to 1.09), indicating systematic discrepancies between methods. Pelvic rotation presented the lowest agreement (ICC < 0.38), with non-significant differences and small effect sizes (d = 0.14 to 0.16), suggesting random variability and poor consistency. These results suggest that the G-Walk provides reasonable estimates for certain pelvic parameters. However, rotation measurements exhibit greater variability, likely due to inconsistencies in sensor placement at L5, soft tissue motion, and the limitations of a single-sensor setup compared to the multi-camera reference system of the GaitLab (Ferrari et al., 2016). Studies like Bugané et al. (2014) found acceptable accuracy in pelvic obliquity using inertial sensors in healthy adults, while Salarian et al. (2004) reported challenges in capturing dynamic rotations in Parkinson’s patients, aligning with our observations. The substantial variability in pelvic angles (CV > 27%) indicates that, even with moderate agreement, the G-Walk’s precision for angular measurements, especially rotation, remains limited, requiring caution in their interpretation.

The sample size of 59 participants is a notable strength of this study, surpassing typical validations of inertial sensors against 3D systems, which often range from 7 to 23 subjects (Hartmann et al., 2009). Additionally, the statistical power was sufficient to detect moderate effect sizes (Cohen’s d = 0.5) with 95% confidence, reinforcing the credibility of non-significant findings. This robustness, combined with the focus on active older adults and the pioneering inclusion of pelvic angles, positions our work as a unique contribution to literature. Reviews such as Sethi, Bharti & Prakash (2022) note that most validation studies target young or pathological populations, with scarce normative data for active older adults, a group relevant due to the link between physical activity and gait preservation (Boyer, Andriacchi & Beaupre, 2012). The high agreement in spatial parameters suggests that the G-Walk is a viable alternative for rapid assessments in clinical or home settings, as supported by Middleton et al., who emphasize walking speed as a “sixth vital sign” in older adults (Fritz & Lusardi, 2009). However, the low reliability in temporal and angular parameters, such as double support time—crucial for fall risk assessment (Beauchet et al., 2017)—indicates that the G-Walk cannot fully replace the GaitLab in scenarios requiring high precision.

These findings carry practical and theoretical implications. Clinically, the G-Walk can serve as a complementary tool for monitoring functionality in active older adults, though methodological enhancements, such as higher sampling rates or refined algorithms, are needed to improve temporal and angular accuracy, as proposed by Ferrari et al. (2016). Theoretically, our results advance the understanding of inertial sensors’ strengths and limitations, aligning with reviews like Iosa et al. (2016) which highlight their efficacy in global measures but challenges in fine details. This study has some limitations. The lack of subgroup analysis (e.g., by sex or age range), which may mask gait pattern variations, as noted by Aboutorabi et al. (2016) The sample was composed exclusively of active older adults, which may limit the generalization of results to sedentary or frail individuals (Dewolf et al., 2021). Another limitation is the unbalanced sex distribution, with a predominance of female participants. However, since the objective was to validate the agreement between two measurement systems through within-subject comparisons, and not to analyze sex-specific gait characteristics, this imbalance does not compromise the integrity of the findings. Future research could address these gaps and explore technological improvements in the G-Walk to broaden its applicability.

Conclusions

In conclusion, the BTS G-Walk proved to be a valid tool for measuring spatial gait parameters in active older adults, with moderate agreement in some pelvic angles, but showed limitations in temporal parameters and pelvic rotation due to differences in sampling frequency, algorithms, and sensor placement. With a robust sample, adequate statistical power,C and the novel inclusion of pelvic angles, this study underscores the G-Walk’s clinical potential while pointing to the need for refinements to enhance its precision in gerontological applications.

Supplemental Information