{"paper_id":"2e199af2-0c62-417a-b4d6-a2c31ff2ceb9","body_text":"Motor and Balance Assessment in Multiple Sclerosis Using Augmented Reality: Concurrent Validation Against Laboratory Reference Measures | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Motor and Balance Assessment in Multiple Sclerosis Using Augmented Reality: Concurrent Validation Against Laboratory Reference Measures Yuri Russo, Edward Nyman Jr, Agne Straukiene, Elliot Winch, Jiaxi Ye, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9424668/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 5 You are reading this latest preprint version Abstract Background: Multiple sclerosis (MS) commonly affects balance and mobility, yet routine clinical tests may lack sensitivity to subtle change. Standardized mobility assessments such as the Timed Up and Go (TUG), six-minute walk test (6MWT), and Romberg balance test are widely used in neurorehabilitation but rely on manual timing or laboratory instrumentation. In this study we evaluated whether motor outcomes derived from augmented-reality (AR) glasses (Magic Leap 2) can provide valid, minimally instrumented measures of balance and gait. Methods: Forty-six adults (26 people with MS; 20 healthy controls, HC) completed Romberg balance testing, TUG, and 6MWT while wearing a head-mounted AR device. AR outcomes were derived from AR glasses and processed using proprietary Strolll algorithms. Gold-standard reference measures were collected concurrently using 18-camera motion capture (100 Hz) and force plates for Romberg (1000 Hz). Primary outcomes were sway area (Romberg; 95% ellipse area), task duration and total distance (TUG/6MWT); secondary outcomes included mean gait velocity and step count (TUG/6MWT) as well as cadence, step length, and number of laps (6MWT). Concurrent validity was quantified using ICC(2,1) absolute agreement, Bland-Altman plots, mean absolute error and root mean square error. Results: Agreement was excellent for primary outcomes, including TUG duration (ICC = 0.927–0.992) and 6MWT total distance (ICC = 0.966), with low absolute error relative to clinical variability. Secondary gait-derived metrics demonstrated good-to-excellent agreement (ICC range 0.750–0.980). AR-derived sway area showed poor-to-good agreement with force plate COP sway area across stance conditions. Agreement was preserved across healthy and clinical cohorts and across task variations. Conclusions: Head-mounted AR–derived outcomes demonstrated good-to-excellent concurrent validity against laboratory gold standards in MS and controls, with strongest performance for global mobility metrics (i.e., duration, distance, mean velocity) and more variable performance for step-derived measures (e.g. cadence, step length). These findings support AR-based assessment as a valid minimally instrumented approach, providing measurement performance consistent with gold-standard metrics for TUG and 6MWT (time, distance) and showing its potential for objectively monitoring disease progression and supporting rehabilitation in clinical and community settings. multiple sclerosis augmented reality digital biomarkers gait analysis wearable sensors home assessment Romberg TUG 6-minute walk test Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 INTRODUCTION Multiple sclerosis (MS) is a chronic neurological condition frequently associated with gait, balance, and non-motor impairments that evolve across the disease course and contribute substantially to functional limitations, mobility loss, and reduced quality of life [ 1 – 4 ]. Subtle deterioration in postural control and movement coordination can appear from the earliest clinically recognised stages of MS (and often before diagnosis), even when traditional clinical scales (e.g., EDSS, 9HPT, T25FW [ 5 , 6 ]) remain stable [ 7 – 9 ]. This can be partly explained by the fact that commonly used clinical assessments of motor function in MS are designed primarily for broad disability staging and therefore offer limited granularity, relying on coarse scoring systems (often with significant inter-rater variability) that are not well suited to tracking small but meaningful within-person changes in motor output [ 10 ]. These limitations are further compounded by clinical assessments that depend on brief, clinic-based observation, non-standardised procedures, and subjective and categorical scoring [ 5 , 6 ]. Finally, infrequent clinical evaluations can often miss short-lived or fluctuating signs, as they provide sparse sampling of a disease course characterised by day-to-day variability, relapses (especially relevant in the most common form of MS), and treatment-related change; as a result, clinically relevant change may go unrecognised between visits and important therapeutic decisions may be delayed. Unlike fully immersive virtual reality, augmented reality (AR) preserves natural vision and real-world perception, making it better suited for balance and gait assessments where safety and more ecologically representative behaviours are essential. Specifically, AR technologies offer an opportunity to overcome the existing limitations of traditional MS scales by enabling standardised, flexible, and highly reproducible motor assessments that can be performed not only in clinical spaces but also in community settings [ 11 , 12 ]. Further, current-generation AR glasses provide multimodal sensing capabilities (e.g. accelerometer, gyroscope, eye tracking), real-time spatial mapping, and controlled visual cues that can be deployed in laboratory as well as home settings for both monitoring and training purposes [ 12 – 14 ]. Project PARAMS (Care closer to home: motor-function Parameters from Augmented-Reality-supported Assessments in people living with Multiple Sclerosis) was developed through a partnership between Torbay and South Devon NHS Foundation Trust, Strolll, and the University of Exeter, with the goal of delivering an AR-enabled platform capable of supporting digital assessment of balance and gait, in people with MS. The proposed assessment uses AR glasses (Magic Leap 2) combined with Strolll’s proprietary algorithms to extract spatiotemporal metrics from motor tasks widely used in clinical assessments of posture and gait. However, before Strolll’s software outputs can support clinical monitoring or remote assessment, technical validation against gold-standard instrumentation is required. The tasks included in the PARAMS study were developed with input from a steering group including people with a diagnosis of MS and clinicians. While the main structure for each task was selected based on clinical relevance, details of how the user, clinician and technology interact to produce the necessary calibrations and settings for specific conditions were largely based on recommendations from stakeholders. Wearable inertial measurement units (IMUs) have emerged as an accessible and relatively low-cost option for capturing gait and balance parameters, and several systems have shown promise for use in MS assessment and other neurodegenerative diseases [ 15 – 17 ]. However, IMU-based approaches remain constrained by various sources of error, including sensitivity to sensor placement, magnetic disturbance, drift, and reduced precision especially when worn for long durations [ 18 – 20 ]. AR-based assessments such as those used in PARAMS could eventually help to mitigate these limitations by leveraging integrated eye-tracking and spatial-mapping capabilities. Eye tracking and spatial-mapping provide high-frequency information on visual attention and gaze behaviour that could potentially support more accurate kinematic estimation and reduce reliance on inertial signals alone [ 21 , 22 ]. Eye tracking, specifically, can be useful for verifying adherence to task instructions, for example by confirming whether participants maintain visual fixation or attend to specified visual targets during the assessment. Furthermore, the AR environment enables the safe presentation of interactive digital objects, supporting exergaming-style tasks that enhance engagement without introducing physical hazards for individuals with balance or mobility impairments [ 23 , 24 ]. Together, these features may offer distinct advantages over stand-alone IMU systems by combining richer sensing modalities with standardised and adaptive tasks that are designed to deliver meaningful measurements while prioritising safety, particularly regarding the avoidance of task-irrelevant hazards. The present study aimed to evaluate the concurrent validity of Strolll motor outcomes of widely used clinical assessments of balance and gait against gold-standard motion capture and force plate measurements. We tested the Strolll algorithms across a set of functional tasks commonly used in MS assessment, including standing balance, transitional mobility, and sustained walking. By comparing AR-derived outcomes with reference biomechanical data in both people with MS and healthy controls, this study provides the first comprehensive technical validation of these AR-based motor assessments in this population. Establishing measurement validity and accuracy is an essential step towards future clinical adoption, remote monitoring applications, and integration of AR-based assessments within MS care pathways. METHODS Participants A total of 46 adults took part in the study, including 26 people with multiple sclerosis (MS) and 20 healthy controls (HC). Participants with MS were recruited through the local community, clinical networks, and social media adverts between March and September 2025. Eligibility criteria for the MS group included a neurologist-confirmed diagnosis, age 18 years or older, and the ability to walk independently for at least one minute with or without a walking aid. Participants were excluded if they had experienced a relapse or major medication change within the previous 30 days, or if they presented with comorbid neurological, psychiatric, or musculoskeletal conditions that could influence balance or gait. Individuals with uncorrected visual impairments or moderate-to-severe cognitive impairment (Mini-Cog < 3 [ 25 ]) were also excluded. Healthy controls were required to have no history of neurological, vestibular, or mobility-affecting musculoskeletal conditions. Individuals who normally wore glasses for walking were asked to use contact lenses during testing to prevent visual obstruction inside the augmented-reality glasses. The study was approved by the University of Exeter Public Health and Sports Sciences Research Ethics Committee (IRB: 6551385), and all participants provided written informed consent prior to participation. Procedures Testing took place during a single session at the Biomechanics Laboratory at St Luke’s Campus, University of Exeter. Before participant arrival, the laboratory space was scanned using the built-in Magic Leap 2 spatial-mapping routine to allow stable placement of AR objects and reliable operation of its simultaneous localisation and mapping. Participants were then fitted with the augmented-reality glasses, adjusting the rear strap for a secure fit and the compute unit was attached either to their trousers' waistband or to a belt provided by the experimenters posteriorly at the L5 level. Passive reflective markers were placed on anatomical landmarks following the Plug-In-Gait protocol [ 26 ], with minor modifications to accommodate wearing AR glasses and its compute unit. Markers were placed on the feet (toe, calcaneus, lateral malleolus), knees (lateral condyle), pelvis (bilateral anterior superior iliac spines and compute pack as a proxy for posterior iliac spines), and on the AR glasses to capture head motions (2 on each side). Participants were familiarised with both the weight of the AR glasses and the augmented-reality environment. The anthropometric calibration procedure was then completed to derive participant-specific parameters such as standing height, seated height, and functional limb lengths. Participants completed three motor tasks presented through the AR interface: a standing balance assessment (Romberg test), the Timed Up and Go (TUG) test, and the Six-Minute Walk Test (6MWT). All tasks were performed while motion-capture cameras and force plates synchronously recorded biomechanical data. Participants were allowed to rest between tasks and a researcher remained alongside them at all times during testing and breaks to ensure safety. The Romberg task included eight 30-second trials featuring different combinations of visual condition (eyes open or closed), stance (feet together or tandem), and support surface (firm or foam). Visual cues for the eyes-open conditions were provided by an AR red balloon (Fig. 1 panel A). For tandem stance, the participant’s preferred forward foot was selected during the familiarisation and maintained throughout the rest of the tandem conditions. Participants who used walking aids were permitted to use them during firm-surface trials but were not asked to perform the foam-surface trials. Experimenters ensured that the walking aids were fully on the force plate. The TUG task required participants to stand from a chair without arms, walk three metres toward an AR pole, turn around it, and return to sit (Fig. 1 , panel B). They were instructed that they could walk around the cone in either direction. A verbal “GO” prompt initiated each trial. Both single-task and dual-task conditions were performed. In the dual-task condition, participants completed a verbal Stroop task in which the words “high” or “low” were spoken in either a high or low pitch. Participants were required to repeat the pitch they heard, regardless of word meaning, and were familiarised with this task before performing the dual-task TUG. The 6MWT required participants to walk continuously for six minutes along a figure-of-eight path defined by two virtual poles placed 3 metres apart (Fig. 1 , panel C). Participants could use their usual walking aid if required. Participants who wanted to stop before the end of the trial were instructed to stop and stand still, until the experimenter stopped the recording. Data Analysis and Outcomes All laboratory reference data were collected using an 18-camera optical motion-capture system (Vicon, UK, 100 Hz) and, for the Romberg test, strain-gauge force plates (AMTI, 1000 Hz). Motion-capture trajectories and force-plate data were synchronised in Vicon Nexus. Small trajectory gaps shorter than 100 ms were linearly interpolated; longer gaps were reconstructed using rigid-body and kinematic fitting algorithms within Nexus. Kinematic signals were then processed in MATLAB (MathWorks, USA) using the BTK Toolkit functions. Head position and pelvis position were computed as the centroid of four head or pelvis markers. Vertical displacement of heel markers (left and right) was used to detect step events. Marker data were sampled low-pass filtered using a 4th-order zero-lag Butterworth filter (15 Hz, [ 27 ]). To allow direct comparison with AR-derived metrics, trial start was defined identically for both systems in each task, ensuring consistent temporal alignment (see each test for further details). 1. Romberg Centre-of-pressure (COP) trajectories (mediolateral and anteroposterior) were estimated from ground-reaction forces and moments. COP data were low-pass filtered using a 4th-order, zero-lag Butterworth filter with a 10 Hz cut-off [ 28 ]. At the start and end of each trial, the experimenter stepped onto a separate force plate that was hardware-synchronised with the main force-plate system while following the countdown displayed in the AR glasses. Synchronisation with the AR system was performed offline by identifying the two vertical-force peaks recorded by the secondary force plate (threshold: 50 N; minimum temporal distance: 15 s), using the second peak as the alignment event. A 29-s window immediately preceding this event was analysed for the test. Sway was quantified using the 95% confidence ellipse area [for a definition see [ 29 ]]. COP trajectories were mean-centred, the 2×2 covariance matrix computed, and principal axes derived via eigen-decomposition. The ellipse was scaled using the chi-square value for two degrees of freedom (χ² = 5.991), and the resulting area (mm²) was used as the primary outcome of the test. AR sway area estimates were derived from head-motion sensors in the AR glasses. Because these measurements reflect head displacement rather than COP motion measured by the force plates, a calibration step was applied to map AR-derived sway areas into the COP measurement space. Calibration was performed using a log–log linear regression across all paired trials (force plate vs. AR raw sway area) using the following formula: log10(force plate sway area) = 0.607 × log10(AR raw sway area) + 1.168 This regression estimated the relationship between the two measurement modalities rather than scaling values to a population average. Because the regression was derived across all trials and produced a near-linear relationship between modalities, the calibration implicitly accounted for differences in sensor location and postural control strategies while preserving trial-to-trial variation in sway magnitude. The resulting transformation was then applied to all AR sway estimates to produce calibrated sway area values used in the final analysis. 2. Timed Up and Go (TUG) The start of the task was identified using vertical head displacement: a baseline was computed from the first 3 seconds of seated data, and task onset was defined as the first frame at which head height exceeded this baseline by more than 3 SD. Task end was defined as the first frame at which head height returned to baseline within the same threshold for at least 3 consecutive seconds. Automated segmentation was visually inspected and manually corrected when required. For each trial, 4 outcomes were extracted: (1) task duration, defined as the interval between detected start and end frames; (2) total number of steps; (3) mean gait velocity, as the cumulative displacement of the head during the task divided by task duration; (4) total distance. Task duration and total distance were considered primary outcomes, whereas total number of steps and mean gait velocity secondary outcomes. 3. Six-Minute Walk Test (6MWT) The 6MWT was analysed using head and heel trajectories. Trial start was detected from forward head velocity: walking commenced when absolute head velocity exceeded baseline (first 3 s) by > 3 SD. The end of the trial was identified when velocity returned below this threshold. Automated segmentation was visually inspected and manually corrected when required. Summary outputs for the test included the same 4 outcomes previously described for the TUG test. Additionally, cadence, step length and number of laps were estimated. First, to identify laps, the 2D head trajectory was projected onto its first principal component, and alternating changes in sign of smoothed velocity were used to detect turning points. Consecutive turn pairs defined individual laps. Task duration and total distance were considered primary outcomes, whereas total number of steps, average step length, cadence, number of laps and mean gait velocity secondary outcomes. AR Data Processing Data from the augmented-reality glasses were processed using Strolll’s proprietary algorithms embedded within their software code. All outcomes for both AR and gold-standard datasets were exported in standardised formats to ensure reproducibility and comparability across tasks. To support transparency and generalisability, Strolll’s algorithms used for AR-based outcome generation are provided as pseudocode in the study data repository (see Availability of Data and Materials). Statistical Analysis Outlier handling was task-specific and designed to preserve physiologically plausible variability while excluding confirmed measurement failures. Differences between the gold-standard and AR measurements were converted to Z-scores. Observations exceeding ± 3 standard deviations of the difference distribution were flagged as potential outliers and subsequently reviewed to confirm that they were not attributable to system error (e.g., hardware malfunction, premature trial termination, segmentation failure). Trials affected by confirmed system errors were excluded from analysis. If no device failure was identified, flagged observations were retained in the dataset and analysed. Concurrent validity between AR-derived outcomes and gold-standard biomechanical measurements was assessed for each task. Relative agreement between the two measurement systems were examined using intraclass correlation coefficients (ICC two-way random effects, absolute agreement, single measure). ICC values were interpreted according to Koo & Li[ 30 ]: <0.50 poor, 0.50–0.75 moderate, 0.75–0.90 good, and > 0.90 excellent agreement. For 6MWT duration, ICCs were not computed because all participants completed the assessment, resulting in a restricted range and insufficient between-subject variability for a meaningful ICC estimate. Absolute agreement was evaluated using Bland–Altman plots, including mean bias and 95% limits of agreement. In these plots, the mean bias (shown as a red line) reflects the average systematic difference between methods, whereas the 95% limits of agreement (shown as dashed lines) indicate the range within which approximately 95% of paired differences are expected to lie. Bias close to zero, narrower limits of agreement, and the absence of an obvious trend across the measurement range were considered indicative of better agreement, whereas a trend in differences with increasing magnitude suggested proportional bias[ 31 ]. Mean and standard deviation values for both the gold-standard and AR measures were also reported, together with mean absolute error (MAE) and root mean square error (RMSE). Analyses were conducted for the full sample and separately for the MS and HC groups (except for the Romberg test, due to the limited sample size). Level of significance was set at p < 0.05. All analyses were performed in custom scripts written in MATLAB (Mathworks, R2023b). Results Twenty healthy controls and twenty-six participants with a diagnosis of MS were tested. Three participants with MS and one HC were excluded from the analysis due to technical issues during data collection. People with MS had a mean age of 54.1 (± 9.4) years, while HC had a mean age of 44.7 (± 13.5) years. All people with MS had a diagnosis of Relapsing-Remitting MS with the exception of two that had Primary-Progressive MS and one that had Secondary-Progressive MS. The average number of years since diagnosis was 10.6 (± 6.5). Three people with MS preferred to use walking aids during the walking task. Further, due to balance capacities, only 7 people with MS and 16 HC completed successfully the foam subtasks of the Romberg test. Romberg Two hundred and thirty-three paired balance trials were included in the analysis. Twenty-seven trials were discarded due to technical issues with the recordings. The most common problems were failure of the AR device, failure of the force-plate, and missing synchronisation. Absolute agreement between calibrated AR sway area and force-plate COP sway area was subtask dependent, ranging from poor (feet together, eyes open on foam condition) to good (tandem, eyes open on firm surface condition). Results divided for sub-conditions are reported in Table 1 , Bland-Altman plots of agreement are reported in Fig. 2 . Results stratified for both subtask and group are provided in the supplementary materials (Table S1). Table 1 Validation results for Romberg test outcomes across the different conditions. The table reports the condition, number of paired observations (force plates vs AR), intraclass correlation coefficient (ICC) with 95% confidence intervals, mean, standard deviation (SD) and agreement metrics (mean absolute error, MAE; root mean square error, RMSE). Condition Number of Pairs ICC(A,1) ICC 95% CI Mean ± SD (mm 2 ) (Force plate) Mean ± SD (mm 2 ) (AR) MAE RMSE Feet Together Floor Eyes Open (mm 2 ) 36 .723 .185 − .756 707 ± 548 680 ± 374 418 650 Feet Together Floor Eyes Closed (mm 2 ) 36 .534 .028 − .775 1319 ± 1299 1067 ± 951 961 1775 Feet Together Foam Eyes Open (mm 2 ) 23 .488 − .061 − .645 900 ± 461 824 ± 583 737 958 Feet Together Foam Eyes Closed (mm 2 ) 22 .443 .142 − .775 1713 ± 1021 1846 ± 3899 797 1084 Tandem Floor Eyes Open (mm 2 ) 35 .671 .719 − .936 896 ± 1015 1323 ± 1008 441 750 Tandem Floor Eyes Closed (mm 2 ) 35 .505 .305 − .808 8820 ± 10441 5216 ± 4852 6273 12226 Tandem Foam Eyes Open (mm 2 ) 23 .936 .400 − .869 1353 ± 2155 1814 ± 2553 479 693 Tandem Foam Eyes Closed (mm 2 ) 23 .615 .239 − .862 9296 ± 10175 6376 ± 5487 6817 10977 Figure 2 here. Figure 2 . Bland–Altman plot for Romberg sway area. The plot shows agreement between force-plate–derived centre-of-pressure (CoP) sway area and the corresponding AR-derived sway area across Romberg conditions. The y-axis represents the difference between methods (Force plates − AR) and the x-axis represents the mean of the two methods. The solid red line indicates the mean bias, and the dashed grey lines indicate the 95% limits of agreement (mean bias ± 1.96 SD). Each point represents an individual trial. Panel A shows the feet together subtasks, while Panel B shows the Tandem subtasks. Timed Up and Go (TUG) A total of 76 matched pairs were included in the final analysis. Six trials were excluded, four due to marker occlusion and two due to AR system failure. Detailed results of the group analysis (HC: 19; MS: 22) are reported in Table 2 , whereas Bland-Altman plots of agreement are reported in Fig. 3 . Table 2 Validation results for Timed Up and Go (TUG) outcomes. Single-task (ST) and dual-task (DT) trials were pooled. For healthy controls (HC) and participants with multiple sclerosis (MS), the table reports agreement for each outcome (duration, mean gait velocity, total distance, and number of steps), quantified using ICC with 95% confidence intervals, mean, standard deviation (SD), mean absolute error (MAE), and root mean square error (RMSE). Metric Group ICC(A,1) ICC 95% CI Mean ± SD (Motion Capture) Mean ± SD (AR) MAE RMSE Duration (s) HC .975 [.951 − .988] 11.3 ± 4.0 12.8 ± 4.4 .32 .88 MS .954 [.917 − .975] 15.9 ± 7.3 17.3 ± 7.7 .56 2.25 Mean Gait Velocity (m/s) HC .869 [.741 − .935] .59 ± .07 .75 ± .15 .05 .07 MS .903 [.818 − .948] .57 ± .14 .59 ± .15 .03 .06 Total Distance (m) HC .989 [.942 − .996] 7.9 ± 2.0 7.7 ± 1.9 .19 .30 MS .943 [.887 − .971] 8.1 ± 1.2 7.9 ± 1.2 .25 .40 Total Steps (number) HC .863 [.734 − .931] 15.2 ± 5.2 14.2 ± 5.3 1.74 2.81 MS .701 [.443 − .841] 20.6 ± 8.1 17.8 ± 6.6 3.88 5.96 Figure 3 here Figure 3 . Bland–Altman plots for Timed Up and Go (TUG) outcomes. Panels show comparisons between motion-capture–derived reference outcomes and AR-derived metrics for TUG measures (i.e., duration, mean gait velocity, total distance and step count). For each panel, the y-axis represents the difference between methods (motion capture - AR) and the x-axis the mean of the two methods. The solid red line indicates the mean bias, and the dashed grey lines indicate the 95% limits of agreement (mean bias ± 1.96 SD). Each point represents an individual trial. Six-minute walk test (6MWT) A total of 42 matched pairs were included in the final analysis. Detailed results of the group analysis (HC: 19; MS: 23) are reported in Table 3 , whereas Bland-Altman plots of agreements are reported in Fig. 4 and Fig. 5 . Table 3 Validation results for Six-Minute Walk Test (6MWT) outcomes. Results are presented separately for healthy controls (HC) and participants with multiple sclerosis (MS). For each outcome (duration, mean gait velocity, total distance, total steps, number of laps, cadence, and average step length), the table reports agreement quantified using ICC with 95% confidence intervals, mean, standard deviation (SD), mean absolute error (MAE), and root mean square error (RMSE). Metric Group ICC(A,1) ICC 95% CI Mean ± SD (Motion Capture) Mean ± SD (AR) MAE RMSE Duration (s) HC NP - 360.0 ± .2 358.9 ± .7 1.04 1.23 MS NP - 359.9 ± .3 359.0 ± .51 .917 1.06 Mean Gait Velocity (m/s) HC .993 .980-.997 .89 ± .11 .89 ± .12 .012 .014 MS .965 .921 − .985 .71 ± .15 .72 ± .17 .022 .044 Total Distance (m) HC .994 .983-.998 319.8 ± 40.0 318.9 ± 41.0 4.03 4.77 MS .964 .919- .990 255.0 ± 55.5 257.1 ± 60.6 7.97 15.86 Total Steps (number) HC .741 .336 − .904 589.2 ± 68.7 564.0 ± 46.7 32.41 45.19 MS .860 .680 − .940 566.6 ± 80.0 550.7 ± 59.5 29.57 38.54 Number of laps (number) HC .881 .707 − .955 33.5 ± 4.9 32.7 ± 5.6 1.47 2.65 MS .815 .617 - .917 25.2 ± 7.2 23.8 ± 9.3 2.74 5.17 Cadence (steps/min) HC .742 .343 − .904 98.2 ± 11.5 94.1 ± 7.8 5.39 7.52 MS .861 .688 − .940 94.4 ± 13.3 91.9 ± 10.0 4.93 6.40 Average Step Length (m) HC .580 .117 − .829 .54 ± .04 .57 ± .05 .027 .041 MS .900 .656 − .963 .45 ± .07 .47 ± .07 .025 .034 Figures 4 and 5 Here Figure 4 . Bland–Altman plots for 6MWT outcomes. Panels show agreement between motion-capture–derived reference and AR-derived measures outcomes for primary 6MWT metrics (i.e., duration and distance). For each panel, the y-axis represents the difference between methods (Motion capture − AR) and the x-axis represents the mean of the two methods. The solid red line indicates the mean bias, and the dashed grey lines indicate the 95% limits of agreement (mean bias ± 1.96 SD). Each point represents an individual trial. Figure 5 . Bland–Altman plots for 6MWT outcomes. Panels show agreement between motion-capture–derived reference and AR-derived measures outcomes for secondary 6MWT metrics (i.e., total steps, number of laps, cadence, and average step length). For each panel, the y-axis represents the difference between methods (Motion capture − AR) and the x-axis represents the mean of the two methods. The solid red line indicates the mean bias, and the dashed grey lines indicate the 95% limits of agreement (mean bias ± 1.96 SD). Each point represents an individual trial. Discussion This study evaluated the concurrent validity of motor and balance outcomes derived from AR glasses, and analysed through Strolll proprietary algorithms, against gold-standard laboratory instrumentation (i.e., motion capture and force plates) in people with MS and age-matched HC. Overall, the AR derived metrics demonstrated task-dependent validity: agreement was poor-to-excellent for static balance (Romberg sway area), and moderate-to-excellent for key functional mobility and walking outcomes (Timed Up and Go; Six-Minute Walk Test). Particularly, agreement was generally strongest for global spatiotemporal parameters (e.g., task duration, mean gait velocity, total distance), whereas step-derived outcomes (e.g., step count, cadence, step length) showed more variable performance. Importantly, the strongest-performing outcomes in this validation correspond to the same primary metrics commonly obtained “by hand” by clinicians in routine clinical practice (e.g., TUG duration, 6MWT total distance). The good-to-excellent agreement observed for these global mobility metrics indicates that the AR-derived estimates provide quantitatively comparable values suitable for assessment and longitudinal monitoring and for early and subtle neurological changes. Overall, the pattern observed here - high agreement for global mobility outcomes and more variable agreement for step-level metrics - is consistent with broader validation work in AR and wearable mobility assessment [ 32 , 33 ]. For example, prior validation studies have reported excellent agreement for TUG completion duration and temporal gait outcomes, but weaker performance for some turning- or spatial measures of gait [ 34 , 35 ]. This aligns with the idea that event detection (e.g., discrete step events) and derived measures that depend on accurate segmentation may be more sensitive to algorithm selection and signal quality than continuous measures, and that wearable sensors tend to be less performative for spatial outcomes [ 36 ]. The comparatively lower agreement for step-derived outcomes in parts of our dataset likely reflects several interacting factors: first, step detection definitions differ by system (heel-marker events in motion capture vs AR-based estimation), which can introduce small discrepancies that accumulate over time (especially in long tasks such as the 6MWT); second, ICC is sensitive to between-subject variability: when a cohort is relatively homogeneous on a given metric, ICC coefficients can appear lower even when absolute errors are small. This is consistent with our observation that some HC step-derived ICCs were modest despite relatively small MAE/RMSE (see Table 1 – 3 ). In the following sections, we discuss these task-specific findings in more detail (Romberg, TUG, and 6MWT), and consider their implications for objective monitoring and technology-enabled rehabilitation in MS. Romberg For quiet-stance balance (i.e., Romberg), calibrated AR sway area demonstrated poor-to-good agreement with force-plate CoP sway area. Agreement varied by condition, ranging from lower values in some feet-together foam conditions to high agreement in tandem stance on firm ground (see Table 1 ). We interpret these differences as arising from three main factors. First, postural control strategies differ substantially across stance configurations, particularly in tandem stance. Compared with feet-together standing - where sway is often dominated by an ankle strategy and primarily expressed in the anteroposterior direction - tandem stance constrains the base of support and increases the challenge to mediolateral stability [ 37 , 38 ]. Participants may adopt a stiffer control strategy (e.g., increased co-contraction and greater reliance on hip/trunk contributions), with larger mediolateral excursions and a higher likelihood of “side loss of balance”. These strategy changes may increase the coupling between global body motion and head kinematics, thereby improving alignment between AR head-based sway estimates and CoP-derived outcomes in some tandem conditions. Second, Romberg conditions performed on foam introduce a compliant interface between the feet and force plate. This manipulation both alters sensory weighting (e.g., reduced reliable somatosensory/proprioceptive input) and changes the mechanics at the foot–support interface, typically resulting in increased CoP excursion and variability. In addition, foam can introduce components in the ground reaction force that reflect surface deformation dynamics (e.g., compression–rebound behaviour and shear effects) rather than postural control alone. These factors weaken the direct relationship between CoP and centre-of-mass (CoM) motion assumed by inverted-pendulum models [ 37 , 38 ], and may therefore reduce correspondence between CoP-based metrics and head-based kinematic estimates. Third, the two outcomes are related but not identical: force-plate-based posturography quantifies control at the base of support (i.e., CoP), whereas AR estimates here are derived from head-based sensing. This distinction was intentional. While lumbar-mounted inertial measurement units are widely used to estimate balance in both research and clinical contexts [ 36 ], we specifically evaluated whether head-based sensing could provide valid balance-related information with minimal instrumentation. Further, head kinematics may capture functionally relevant aspects of stability - such as upper-body control and head stabilisation required to maintain gaze and manage vestibular demands - which are important in daily activities and may relate to fall risk [ 39 , 40 ]. Lastly, head-based sensing removes the need for floor-mounted measurement devices. Although low-cost systems can provide meaningful balance metrics under standard conditions, their performance can be compromised on compliant surfaces such as the foam used in the Romberg sub-tests. In this respect, AR-based approaches offer a practical advantage [ 41 ]. Accordingly, we believe that head-based metrics should not replace force-plate posturography/lumbar-based measures, but to provide a complementary perspective - particularly relevant for scalable monitoring and tasks in which upper-body control is a major contributor to functional stability. Importantly, given the cost, set-up time, and limited portability of force-plate systems, head-based measures collected through portable devices, such as the one employed in this study, may also serve as a minimally instrumented option in contexts where laboratory instrumentation is impractical or unavailable, including remote and field-deployable assessments. Timed Up and Go For the TUG test, agreement between the AR-based and the motion capture outcomes was consistently strong across outcomes (both primary and secondary) and groups. Specifically, agreement was excellent for duration and total distance, with good agreement for mean gait velocity and step count (see Table 2 ). The strong agreement for duration is consistent with our processing approach, which defines task onset/offset from vertical head displacement (sit-to-stand and return-to-sit). These findings align with prior instrumented-TUG validation work showing that total time can be captured with very high agreement relative to motion capture systems, while reducing reliance on operator-dependent errors (e.g., stopwatch) [ 42 ]. From a broader rehabilitation and monitoring perspective, the value of instrumenting the TUG is not simply replicating total time but enabling objective quantification of components that reflect balance control and fall risk (e.g., turning and chair transfers). Nevertheless, demonstrating that automated AR-derived duration closely matches laboratory-derived duration supports immediate applicability as a minimally instrumented alternative to stopwatch timing, while enabling richer and more objective subcomponent metrics in future work. Previous work highlighted that conventional TUG scoring can miss intervention effects because it collapses multiple mobility skills into a single time outcome, whereas instrumented approaches enable objective subcomponent analysis (sit-to-stand, gait, turning, turn-to-sit) and richer motor evaluations [ 43 ]. In fact, recent work using IMUs in TUG tasks highlights that turning and sitting-related parameters (e.g., reduced turning angular velocity; altered acceleration during stand-to-sit) can be particularly informative for detecting deterioration in physical function, reinforcing the clinical relevance of subphase and metrics beyond total time [ 44 – 47 ]. Further, incorporating a cognitive dual-task during the TUG (e.g., as in the Mini-BESTest [ 48 ]) provides an opportunity to investigate cognitive–motor interactions, and laboratory measures under dual-task conditions may better reflect everyday mobility demands than single-task assessments [ 49 ]. AR-based systems have already shown that several TUG sub-durations can be derived with excellent between-systems agreement, although turning segmentation can be more challenging than straight walking or chair transitions [ 50 ]. Taken together, these elements position the present findings as an important enabling step: by demonstrating strong validity for global TUG outcomes, the AR approach is well placed to extend toward richer, phase-specific digital biomarkers in future iterations - particularly relevant for technology-supported rehabilitation pathways that aim to scale objective assessment into routine care and home settings. The comparatively lower agreement for step count in MS likely reflects a combination of algorithmic and behavioural factors. First, step detection is inherently sensitive to atypical movement patterns; as mobility becomes more pathological (e.g., slower speed, altered rhythm, disrupted turning), segmentation and threshold-based detection become more challenging using data from wearable devices and biomechanical laboratory reference systems alike. Second, our protocol allowed testing with or without a walking aid, and assistive device use alters the upper-body motion and intermittently constrains natural gait dynamics, which may affect the accuracy of acceleration-derived step features, compared with spatial reference systems such as motion capture or visual SLAM [ 51 ]. Notably, some instrumented-TUG studies exclude participants who use walking aids, which may strengthen statistical accuracy at the cost of clinical generalisability; in contrast, our approach intentionally prioritises inclusion of a wider and more representative clinical cohort, with the trade-off that step-based metrics in walking-aid users should currently be interpreted more cautiously. Future development should therefore focus on improving robustness of step-event detection in the presence of assistive devices and more impaired gait patterns (e.g., adaptive thresholds and/or multi-feature fusion), to better support deployment in heterogeneous rehabilitation cohorts. 6-Minute Walking Test For the Six-Minute Walk Test (6MWT), agreement was excellent for mean gait velocity and total distance and moderate-to-good for laps and step-derived measures (see Table 3 ). We believe that these results contribute to the existing literature, providing a valid, accurate alternative to provide quantitative assessments of this common test. A key contribution stems from the fact that here the 6MWT was implemented as a figure-of-eight trajectory between two AR poles, rather than as a straight corridor walk. This design choice was pragmatic and clinically motivated as a figure-of-eight walking trajectory demands an equal number of turns in the left and right directions, requires substantially less physical space - a common constraint in clinical environments - and reduces the need for participants to navigate busy corridors, thereby limiting external distractions and making it easier for staff to provide close supervision or physical assistance when needed. Further, this layout is also advantageous for potential home-based assessments, where available walkway length is often limited to a few metres. In addition, a turning-rich trajectory may be particularly valuable from a biomechanical perspective because it increases task complexity relative to straight walking, can help reveal walking asymmetries, and better reflects the demands of everyday mobility where turning is frequent and often challenging. These considerations align with the work of Shah et al., who highlight walkway length and turning content as key determinants of 6MWT distance-estimation error, when using wearable devices, noting that turning is likely a principal contributor to measurement discrepancies when estimating distance from wearable sensors [ 50 ]. Despite the high turning content in our protocol (poles separated by 3 m and most times > 25 laps completed per participant), AR-derived distance and mean gait velocity demonstrated excellent agreement with motion capture in both HC and MS, suggesting that the AR system can provide robust estimates even under turning-intensive walking conditions. It is however worth mentioning that, similarly to the TUG task, the residual differences observed in a small subset of participants (see Figs. 4 and 5 ) may partly reflect the use of walking aids (i.e., walkers and walking sticks). Consequently, we believe that 6MWT AR-derived outcomes in cohorts using walking aids should be interpreted with caution and future developments should explicitly address aid-related movement patterns to improve accuracy and generalisability. Technology-enabled MS disease monitoring and rehabilitation Conventional MS motor assessments can be coarse-grained, partly subjective, and infrequent, which limits sensitivity to subtle within-person change and constrains timely clinical decision-making, especially in the remitting-and-relapse MS subtype. The broader vision of project PARAMS is to enable standardised, objective, home-based motor-function assessment using wearable AR glasses, supporting longitudinal monitoring and trend analysis, and ultimately helping services deliver care “closer to home” while improving efficiency and cost-effectiveness. Within that working frame, the present technical validation provides a critical prerequisite as it demonstrates that several clinically interpretable outcomes - particularly duration, mean gait velocity, and distance across common clinical tests - can be captured with high agreement relative to laboratory gold standards. These outcomes are directly relevant to clinical monitoring and rehabilitation because they are commonly used for baseline profiling, goal setting, and tracking response to therapy or disease progression. In addition, this AR platform can support standardisation (consistent task setup and cueing) and may reduce measurement variability introduced by rater-dependent scoring. Limitations The study has five main limitations. First, data were collected in a laboratory context, and performance may differ in true home environments (e.g., lighting conditions, space constraints, connectivity issues). Second, while the present work focuses on concurrent validity, the broader implementation case for pathway change depends on additional evidence (feasibility, usability, reliability, and responsiveness to change). Ongoing studies are assessing these domains. Third, our MS cohort was predominantly relapsing–remitting, with only a small number of participants with progressive phenotypes. As gait and balance impairment are often greater in progressive MS, the validity of AR-derived outcomes in more advanced or progressive disease (and across wider disability levels) requires confirmation in larger, more phenotypically diverse samples. Fourth, the presence of walking aids during the assessment may reduce the overall accuracy of the AR-based metrics, especially those related to step characteristics, therefore future studies should specifically focus on further developing these algorithms to further expand the generalisability of this approach. Finally, because many participants with MS could not complete the foam subtasks of the Romberg test, we were unable to stratify the validation by group for these conditions. Conclusion This study demonstrates that motor outcomes derived from AR glasses and processed with Strolll’s algorithms show task-dependent concurrent validity against laboratory gold standards (motion capture and force plates) in people with MS and age-matched controls. Agreement was strongest for global mobility metrics that are central to routine clinical assessment - particularly TUG duration and distance and 6MWT distance and mean gait velocity - supporting AR-based measurement as a minimally instrumented alternative when laboratory systems are impractical. Step-derived outcomes showed more variable performance, especially in MS, indicating a priority area for algorithm refinement and robustness to impaired gait patterns and walking-aid use. Collectively, these results support AR-based assessment as a valid and scalable approach for objective motor measurement in neurorehabilitation, and performance in real-world clinical and community settings. Declarations Ethics approval and consent to participate The study was approved by the University of Exeter Public Health and Sports Sciences Research Ethics Committee (IRB: 6551385), and all participants provided written informed consent prior to participation. Consent for publication Not applicable. Availability of data and materials To ensure methodological transparency and reproducibility, motion capture and force plates data as well as all custom MATLAB scripts/functions used for processing and analysing the force-plate and motion-capture datasets (i.e., data import, filtering, segmentation, COP computation, step detection, and gait metric extraction) are available in the Open Science Framework repository: https://osf.io/fup6j . These shared scripts/functions exclude the proprietary Strolll algorithms used for AR-based outcome generation, which are however shared in the form of pseudocode to ensure proper generalization, and therein negating a ‘black-box’ approach, without sharing proprietary code, in the same repository. For the purpose of open access, the author has applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising from this submission. Competing interests E.N. is VP of Data and AI, M.R. is a scientific advisor, and E.W. is a software developer for Strolll, a provider of digital neurorehabilitation software via AR glasses. The funder had no role in the design of the study, in the collection, analysis, or interpretation of data, in the writing of the manuscript, or in the decision to publish the results. YR is a member of the Editorial Board of Journal of NeuroEngineering and Rehabilitation. YR was not involved in the journal’s peer review process of, or decisions related to, this manuscript. Funding This study was funded by Torbay and South Devon NHS Foundation Trust via an award from NHS England through the NIHR Digital Health Partnership Award (Project PARAMS). and was supported by the National Institute for Health and Care Research (NIHR) Exeter Biomedical Research Centre. The views expressed are those of the authors and do not necessarily reflect those of NHS England, the NIHR, or the Department of Health and Social Care. Authors' contributions Study Conception and design – YR, EN, AS, MR & WY. Acquisition of data – YR & WY. Data curation – YR, JY & PL. Algorithm Development – EN & EW. Study Analysis Software – YR & EN. Formal analysis – YR. Visualisation – YR & EN. Funding – YR, AS & WY. Writing original draft - YR & WY. Writing review & editing - All the authors. Final approval of the completed article - All the authors. Acknowledgements We would like to thank the study participants and those who accompanied them to the laboratory. We also thank people with MS, the MS coordinator, the MS nurse, and the Torbay MS Research Team who attended the PPIE workshops and provided guidance on the development of the software. We are grateful to Prof Stefano Pluchino (University of Cambridge) for study design advice and Dr Antonio Scalfari (Imperial College London) for input on the funding application. Finally, we thank Barri Morgan (Strolll) for facilitating communication between the parties involved in PARAMS. References Reich DS, Lucchinetti CF, Calabresi PA. Multiple Sclerosis. New England Journal of Medicine. Massachusetts Medical Society; 2018;378:169–80. https://doi.org/10.1056/nejmra1401483 Comber L, Galvin R, Coote S. Gait deficits in people with multiple sclerosis: A systematic review and meta-analysis. Gait Posture. Elsevier B.V.; 2017. p. 25–35. https://doi.org/10.1016/j.gaitpost.2016.09.026 Comber L, Sosnoff JJ, Galvin R, Coote S. Postural control deficits in people with Multiple Sclerosis: A systematic review and meta-analysis. Gait Posture. Elsevier B.V.; 2018. p. 445–52. https://doi.org/10.1016/j.gaitpost.2018.02.018 Lane J, Ng HS, Poyser C, Lucas RM, Tremlett H. Multiple sclerosis incidence: A systematic review of change over time by geographical region. Mult. Scler. Relat. Disord. Elsevier B.V.; 2022. https://doi.org/10.1016/j.msard.2022.103932 Uitdehaag BMJ. Clinical outcome measures in multiple sclerosis. In: Goodin D, editor. handbook of clinical neurology, Multiple Sclerosis and Related Disorders. 3rd ed. Elsevier; 2014. Inojosa H, Schriefer D, Ziemssen T. Clinical outcome measures in multiple sclerosis: A review. Autoimmun. Rev. Elsevier B.V.; 2020. https://doi.org/10.1016/j.autrev.2020.102512 Marrie RA, Allegretta M, Barcellos LF, Bebo B, Calabresi PA, Correale J, et al. From the prodromal stage of multiple sclerosis to disease prevention. Nat Rev Neurol. Nature Research; 2022;18:559–72. https://doi.org/10.1038/s41582-022-00686-x Melillo F, Di Sapio A, Martire S, Malentacchi M, Matta M, Bertolotto A. Computerized posturography is more sensitive than clinical Romberg Test in detecting postural control impairment in minimally impaired Multiple Sclerosis patients. Mult Scler Relat Disord. Elsevier B.V.; 2017;14:51–5. https://doi.org/10.1016/j.msard.2017.03.008 Vienne-Jumeau A, Quijoux F, Vidal PP, Ricard D. Wearable inertial sensors provide reliable biomarkers of disease severity in multiple sclerosis: A systematic review and meta-analysis. Ann. Phys. Rehabil. Med. Elsevier Masson SAS; 2020. p. 138–47. https://doi.org/10.1016/j.rehab.2019.07.004 Inojosa H, Schriefer D, Klöditz A, Trentzsch K, Ziemssen T. Balance Testing in Multiple Sclerosis—Improving Neurological Assessment With Static Posturography? Front Neurol. Frontiers Media S.A.; 2020;11. https://doi.org/10.3389/fneur.2020.00135 Fernández-Cañas M, Cano-de-la-Cuerda R, Marcos-Antón S, Onate-Figuérez A. Augmented and Mixed Reality Interventions in People with Multiple Sclerosis: A Systematic Review. Brain Sci. Multidisciplinary Digital Publishing Institute (MDPI); 2025. https://doi.org/10.3390/brainsci15121292 van Bergem JS, van Doorn PF, Hoogendoorn EM, Geerse DJ, Roerdink M. Gait and Balance Assessments with Augmented Reality Glasses in People with Parkinson’s Disease: Concurrent Validity and Test–Retest Reliability. Sensors. Multidisciplinary Digital Publishing Institute (MDPI); 2024;24. https://doi.org/10.3390/s24175485 van Doorn PF, Nyman E, Wishaupt K, van der Krogt MM, Roerdink M. Deriving Motor States and Mobility Metrics from Gamified Augmented Reality Rehabilitation Exercises in People with Parkinson’s Disease. Sensors. Multidisciplinary Digital Publishing Institute (MDPI); 2025;25. https://doi.org/10.3390/s25237172 Palumbo A. Microsoft HoloLens 2 in Medical and Healthcare Context: State of the Art and Future Prospects. Sensors. MDPI; 2022. https://doi.org/10.3390/s22207709 Russo Y, Stuart S, Silva-Batista C, Brumbach B, Vannozzi G, Mancini M. Does visual cueing improve gait initiation in people with Parkinson’s disease? Hum Mov Sci. Elsevier B.V.; 2022;84. https://doi.org/10.1016/j.humov.2022.102970 Voisard C, de l’Escalopier N, Vienne-Jumeau A, Moreau A, Quijoux F, Bompaire F, et al. Innovative multidimensional gait evaluation using IMU in multiple sclerosis: introducing the semiogram. Front Neurol. Frontiers Media SA; 2023;14. https://doi.org/10.3389/fneur.2023.1237162 Brognara L, Palumbo P, Grimm B, Palmerini L. Assessing Gait in Parkinson’s Disease Using Wearable Motion Sensors: A Systematic Review. Diseases. MDPI AG; 2019;7:18. https://doi.org/10.3390/diseases7010018 Rast FM, Labruyère R. Systematic review on the application of wearable inertial sensors to quantify everyday life motor activity in people with mobility impairments. J. Neuroeng. Rehabil. BioMed Central Ltd; 2020. https://doi.org/10.1186/s12984-020-00779-y Robert-Lachaine X, Mecheri H, Larue C, Plamondon A. Effect of local magnetic field disturbances on inertial measurement units accuracy. Appl Ergon. Elsevier Ltd; 2017;63:123–32. https://doi.org/10.1016/j.apergo.2017.04.011 Gu C, Lin W, He X, Zhang L, Zhang M. IMU-based motion capture system for rehabilitation applications: A systematic review. Biomimetic Intelligence and Robotics. Elsevier B.V.; 2023. https://doi.org/10.1016/j.birob.2023.100097 Hübner P, Clintworth K, Liu Q, Weinmann M, Wursthorn S. Evaluation of hololens tracking and depth sensing for indoor mapping applications. Sensors (Switzerland). MDPI AG; 2020;20. https://doi.org/10.3390/s20041021 Soares I, Sousa RB, Petry M, Moreira AP. Accuracy and repeatability tests on hololens 2 and htc vive. Multimodal Technologies and Interaction. MDPI; 2021;5. https://doi.org/10.3390/mti5080047 Calafiore D, Invernizzi M, Ammendolia A, Marotta N, Fortunato F, Paolucci T, et al. Efficacy of Virtual Reality and Exergaming in Improving Balance in Patients With Multiple Sclerosis: A Systematic Review and Meta-Analysis. Front. Neurol. Frontiers Media S.A.; 2021. https://doi.org/10.3389/fneur.2021.773459 Dalmazane M, Gallou-Guyot M, Compagnat M, Magy L, Montcuquet A, Billot M, et al. Effects on gait and balance of home-based active video game interventions in persons with multiple sclerosis: A systematic review. Mult. Scler. Relat. Disord. Elsevier B.V.; 2021. https://doi.org/10.1016/j.msard.2021.102928 Borson S, Scanlan JM, Chen P, Ganguli M. The Mini-Cog as a screen for dementia: Validation in a population-based sample. J Am Geriatr Soc. 2003;51:1451–4. https://doi.org/10.1046/j.1532-5415.2003.51465.x Gutierrez-Farewik EM, Bartonek Å, Saraste H. Comparison and evaluation of two common methods to measure center of mass displacement in three dimensions during gait. Hum Mov Sci. 2006;25:238–56. https://doi.org/10.1016/j.humov.2005.11.001 Russo Y, Berchicci M, Di Russo F, Vannozzi G. How do different movement references influence ERP related to gait initiation? A comparative methods’ assessment. J Neurosci Methods. Elsevier B.V.; 2019;311:95–101. https://doi.org/10.1016/j.jneumeth.2018.10.006 Russo Y, Vannozzi G. Anticipatory postural adjustments in forward and backward single stepping: Task variability and effects of footwear. J Biomech. Elsevier Ltd; 2021;122. https://doi.org/10.1016/j.jbiomech.2021.110442 Quijoux F, Nicolaï A, Chairi I, Bargiotas I, Ricard D, Yelnik A, et al. A review of center of pressure (COP) variables to quantify standing balance in elderly people: Algorithms and open-access code*. Physiol. Rep. American Physiological Society; 2021. https://doi.org/10.14814/phy2.15067 Koo TK, Li MY. A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for Reliability Research. J Chiropr Med. Elsevier USA; 2016;15:155–63. https://doi.org/10.1016/j.jcm.2016.02.012 Bland JM, Altman DG. Measurement STATISTICAL METHODS FOR ASSESSING AGREEMENT BETWEEN TWO METHODS OF CLINICAL MEASUREMENT. The Lancet. 1986;307–10. van Doorn PF, Geerse DJ, van Bergem JS, Hoogendoorn EM, Nyman E, Roerdink M. Gait Parameters Can Be Derived Reliably and Validly from Augmented Reality Glasses in People with Parkinson’s Disease Performing 10-m Walk Tests at Comfortable and Fast Speeds. Sensors. Multidisciplinary Digital Publishing Institute (MDPI); 2025;25. https://doi.org/10.3390/s25041230 Koop MM, Rosenfeldt AB, Owen K, Penko AL, Streicher MC, Albright A, et al. The Microsoft HoloLens 2 Provides Accurate Measures of Gait, Turning, and Functional Mobility in Healthy Adults. Sensors. MDPI; 2022;22. https://doi.org/10.3390/s22052009 Mancini M, Horak FB. Potential of APDM mobility lab for the monitoring of the progression of Parkinson’s disease. Expert Rev. Med. Devices. Taylor and Francis Ltd; 2016. p. 455–62. https://doi.org/10.1586/17434440.2016.1153421 Zampieri C, Salarian A, Carlson-Kuhta P, Aminian K, Nutt JG, Horak FB. The instrumented timed up and go test: Potential outcome measure for disease modifying therapies in Parkinson’s disease. J Neurol Neurosurg Psychiatry. BMJ Publishing Group; 2010;81:171–6. https://doi.org/10.1136/jnnp.2009.173740 Micó-Amigo ME, Bonci T, Paraschiv-Ionescu A, Ullrich M, Kirk C, Soltani A, et al. Assessing real-world gait with digital technology? Validation, insights and recommendations from the Mobilise-D consortium. J Neuroeng Rehabil. BioMed Central Ltd; 2023;20. https://doi.org/10.1186/s12984-023-01198-5 Winter DA. Human balance and posture control during standing and walking. Gait Posture. 1995;3:193–214. Gatev P, Thomas S, Kepple T, Hallett M. Feedforward ankle strategy of balance during quiet stance in adults. Journal of Physiology. 1999. Appiah-Kubi KO, Galgon A, Tierney R, Lauer R, Wright WG. Concurrent vestibular activation and postural training recalibrate somatosensory, vestibular and gaze stabilization processes. PLoS One. Public Library of Science; 2024;19. https://doi.org/10.1371/journal.pone.0292200 Garg H, Dibble LE, Schubert MC, Sibthorp J, Foreman KB, Gappmaier E. Gaze Stability, Dynamic Balance and Participation Deficits in People with Multiple Sclerosis at Fall-Risk. Anatomical Record. Blackwell Publishing Inc.; 2018;301:1852–60. https://doi.org/10.1002/ar.23852 Keune PM, Young WR, Paraskevopoulos IT, Hansen S, Muenssinger J, Oschmann P, et al. Measuring standing balance in multiple sclerosis: Further progress towards an automatic and reliable method in clinical practice. J Neurol Sci. Elsevier B.V.; 2017;379:157–62. https://doi.org/10.1016/j.jns.2017.06.007 Kleiner AFR, Pacifici I, Vagnini A, Camerota F, Celletti C, Stocchi F, et al. Timed Up and Go evaluation with wearable devices: Validation in Parkinson’s disease. J Bodyw Mov Ther. Churchill Livingstone; 2018;22:390–5. https://doi.org/10.1016/j.jbmt.2017.07.006 Salarian A, Horak FB, Zampieri C, Carlson-Kuhta P, Nutt JG, Aminian K. ITUG, a sensitive and reliable measure of mobility. IEEE Transactions on Neural Systems and Rehabilitation Engineering. 2010;18:303–10. https://doi.org/10.1109/TNSRE.2010.2047606 Mellone S, Tacconi C, Chiari L. Validity of a Smartphone-based instrumented Timed Up and Go. Gait Posture. 2012;36:163–5. https://doi.org/10.1016/j.gaitpost.2012.02.006 Szaflik P, Kaczmarczyk A, Zadoń H, Szefler-Derela J, Wasiuk-Zowada D, Nowakowska-Lipiec K, et al. Instrumented Timed Up and Go Test as a Tool to Early Detection of Gait and Functional Mobility Impairments in Multiple Sclerosis. J Clin Med. MDPI AG; 2026;15:679. https://doi.org/10.3390/jcm15020679 Zampieri C, Salarian A, Carlson-Kuhta P, Aminian K, Nutt JG, Horak FB. The instrumented timed up and go test: Potential outcome measure for disease modifying therapies in Parkinson’s disease. J Neurol Neurosurg Psychiatry. BMJ Publishing Group; 2010;81:171–6. https://doi.org/10.1136/jnnp.2009.173740 Caronni A, Picardi M, Scarano S, Malloggi C, Tropea P, Gilardone G, et al. Pay attention: you can fall! The Mini-BESTest scale and the turning duration of the TUG test provide valid balance measures in neurological patients: a prospective study with falls as the balance criterion. Front Neurol. Frontiers Media SA; 2023;14. https://doi.org/10.3389/fneur.2023.1228302 Franchignoni F, Horak F, Godi M, Nardone A, Giordano A. Using psychometric techniques to improve the balance evaluation systems test: The mini-bestest. J Rehabil Med. 2010;42:323–31. https://doi.org/10.2340/16501977-0537 Hillel I, Gazit E, Nieuwboer A, Avanzino L, Rochester L, Cereatti A, et al. Is every-day walking in older adults more analogous to dual-task walking or to usual walking? Elucidating the gaps between gait performance in the lab and during 24/7 monitoring. European Review of Aging and Physical Activity. Springer Verlag; 2019;16. https://doi.org/10.1186/s11556-019-0214-5 Shah V V., Curtze C, Sowalsky K, Arpan I, Mancini M, Carlson-Kuhta P, et al. Inertial Sensor Algorithm to Estimate Walk Distance. Sensors. MDPI; 2022;22. https://doi.org/10.3390/s22031077 Theodorou C, Velisavljevic V, Dyo V, Nonyelu F. Visual SLAM algorithms and their application for AR, mapping, localization and wayfinding. Array. Elsevier B.V.; 2022;15. https://doi.org/10.1016/j.array.2022.100222 Additional Declarations Competing interest reported. E.N. is VP of Data and AI, M.R. is a scientific advisor, and E.W. is a software developer for Strolll, a provider of digital neurorehabilitation software via AR glasses. The funder had no role in the design of the study, in the collection, analysis, or interpretation of data, in the writing of the manuscript, or in the decision to publish the results. YR is a member of the Editorial Board of Journal of NeuroEngineering and Rehabilitation. YR was not involved in the journal’s peer review process of, or decisions related to, this manuscript. Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 28 Apr, 2026 Reviewers invited by journal 28 Apr, 2026 Editor assigned by journal 20 Apr, 2026 Submission checks completed at journal 20 Apr, 2026 First submitted to journal 15 Apr, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {\"props\":{\"pageProps\":{\"initialData\":{\"identity\":\"rs-9424668\",\"acceptedTermsAndConditions\":true,\"allowDirectSubmit\":false,\"archivedVersions\":[],\"articleType\":\"Research Article\",\"associatedPublications\":[],\"authors\":[{\"id\":626394551,\"identity\":\"ea13625e-834d-4331-aa3e-356fb25b2357\",\"order_by\":0,\"name\":\"Yuri Russo\",\"email\":\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA1UlEQVRIiWNgGAWjYBADOWQOM1FajIGYsYEkLYkNRGsxOMD7TPLHH5v0fonk4w8+7mGQ52/gMTbAr4XdTJq3LS13Zs+xxMYZzxgMZxzgMU7Ar4WNTZqx4XDuhuM9hs08BxgYNzDwGB8gpAXosMPp9of5Pzb/OcBgT5QWCR62wwkG7D2MzQwHGBJBWvA6TPIwG7M10C+GM84cM5zZc0AiecZhtmK83uc73sZ4Exhi8vwzkh98+HHAxra/vXmzBD4tCodR+RKEI1K+gYCCUTAKRsEoGAUMAOj/RGNZNkp0AAAAAElFTkSuQmCC\",\"orcid\":\"\",\"institution\":\"University of Exeter\",\"correspondingAuthor\":true,\"prefix\":\"\",\"firstName\":\"Yuri\",\"middleName\":\"\",\"lastName\":\"Russo\",\"suffix\":\"\"},{\"id\":626394552,\"identity\":\"f5305dc3-de6a-4ace-9aa9-e97c41ad1390\",\"order_by\":1,\"name\":\"Edward Nyman Jr\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Strolll Limited\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Edward\",\"middleName\":\"\",\"lastName\":\"Nyman\",\"suffix\":\"Jr\"},{\"id\":626394553,\"identity\":\"e52d1dcb-7f56-4070-8da4-c02af99e8a47\",\"order_by\":2,\"name\":\"Agne Straukiene\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Torbay and South Devon NHS Foundation Trust\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Agne\",\"middleName\":\"\",\"lastName\":\"Straukiene\",\"suffix\":\"\"},{\"id\":626394557,\"identity\":\"2f9b6760-0384-4ab9-857c-9338f6740841\",\"order_by\":3,\"name\":\"Elliot Winch\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Strolll Limited\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Elliot\",\"middleName\":\"\",\"lastName\":\"Winch\",\"suffix\":\"\"},{\"id\":626394559,\"identity\":\"4c61e61f-94bb-49ea-918a-a57b04b3d8b2\",\"order_by\":4,\"name\":\"Jiaxi Ye\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"University of Exeter\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Jiaxi\",\"middleName\":\"\",\"lastName\":\"Ye\",\"suffix\":\"\"},{\"id\":626394560,\"identity\":\"966b2a3b-4e72-4a6e-88d8-09e33945178a\",\"order_by\":5,\"name\":\"Phaedra Leveridge\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"University of Exeter\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Phaedra\",\"middleName\":\"\",\"lastName\":\"Leveridge\",\"suffix\":\"\"},{\"id\":626394561,\"identity\":\"c8272a6a-9677-486e-bcd5-cf5096cce56e\",\"order_by\":6,\"name\":\"Melvyn Roerdink\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Vrije Universiteit Amsterdam\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Melvyn\",\"middleName\":\"\",\"lastName\":\"Roerdink\",\"suffix\":\"\"},{\"id\":626394562,\"identity\":\"314cd8eb-c0f5-4ac6-98da-532c88e72417\",\"order_by\":7,\"name\":\"William Young\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"University of Exeter\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"William\",\"middleName\":\"\",\"lastName\":\"Young\",\"suffix\":\"\"}],\"badges\":[],\"createdAt\":\"2026-04-15 09:45:12\",\"currentVersionCode\":1,\"declarations\":\"\",\"doi\":\"10.21203/rs.3.rs-9424668/v1\",\"doiUrl\":\"https://doi.org/10.21203/rs.3.rs-9424668/v1\",\"draftVersion\":[],\"editorialEvents\":[],\"editorialNote\":\"\",\"failedWorkflow\":false,\"files\":[{\"id\":107536510,\"identity\":\"ec6e6665-3d49-49d6-89f9-4a3c8d172dad\",\"added_by\":\"auto\",\"created_at\":\"2026-04-22 11:27:16\",\"extension\":\"png\",\"order_by\":1,\"title\":\"Figure 1\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":47877,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cem\\u003e\\u003cstrong\\u003eAugmented-reality task environments.\\u003c/strong\\u003e\\u003c/em\\u003e\\u003cem\\u003e Example visualisations of the AR environments created for each task. For display purposes, a grey background is shown to enhance the visibility of the AR elements. (A) Romberg task: eyes-open tandem stance condition. People were asked to stand with one foot in front of the other and to focus their gaze on the red balloon in front of them. (B) Timed Up and Go (TUG) test: participants started seated on a chair aligned with the start line and walked around a digital pole positioned 3 metres away; the green guidance track was used for task instruction only and was removed during the test to avoid distraction. (C) Six-Minute Walk Test (6MWT), experimenter view: participants walked a figure-of-eight path between two poles for 6 minutes (or until they needed to stop); the purple guidance track was shown only during instruction and removed during the test.\\u003c/em\\u003e\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"1.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-9424668/v1/0897008e737603a7da80ea3e.png\"},{\"id\":107536618,\"identity\":\"f679152c-626b-4e13-a8ef-c5db2275dd8d\",\"added_by\":\"auto\",\"created_at\":\"2026-04-22 11:27:34\",\"extension\":\"png\",\"order_by\":2,\"title\":\"Figure 2\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":367283,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eBland–Altman plot for Romberg sway area. The plot shows agreement between force-plate–derived centre-of-pressure (CoP) sway area and the corresponding AR-derived sway area across Romberg conditions. The y-axis represents the difference between methods (Force plates − AR) and the x-axis represents the mean of the two methods. The solid red line indicates the mean bias, and the dashed grey lines indicate the 95% limits of agreement (mean bias ± 1.96 SD). Each point represents an individual trial. Panel A shows the feet together subtasks, while Panel B shows the Tandem subtasks.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"2.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-9424668/v1/5f8940a39877363711f2214a.png\"},{\"id\":107536569,\"identity\":\"b2634e1c-fff6-43fc-8fdd-d7136a9e88cb\",\"added_by\":\"auto\",\"created_at\":\"2026-04-22 11:27:22\",\"extension\":\"png\",\"order_by\":3,\"title\":\"Figure 3\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":84027,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cem\\u003eBland–Altman plots for Timed Up and Go (TUG) outcomes. Panels show comparisons between motion-capture–derived reference outcomes and AR-derived metrics for TUG measures (i.e., duration, mean gait velocity, total distance and step count). For each panel, the y-axis represents the difference between methods (motion capture - AR) and the x-axis the mean of the two methods. The solid red line indicates the mean bias, and the dashed grey lines indicate the 95% limits of agreement (mean bias ± 1.96 SD). Each point represents an individual trial.\\u003c/em\\u003e\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"3.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-9424668/v1/eaba3779469298f8ec209878.png\"},{\"id\":107536568,\"identity\":\"48ae80d3-253d-414e-b8d9-571e1181921d\",\"added_by\":\"auto\",\"created_at\":\"2026-04-22 11:27:22\",\"extension\":\"png\",\"order_by\":4,\"title\":\"Figure 4\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":52792,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cem\\u003eBland–Altman plots for 6MWT outcomes. Panels show agreement between motion-capture–derived reference and AR-derived measures outcomes for primary 6MWT metrics (i.e., duration and distance). For each panel, the y-axis represents the difference between methods (Motion capture − AR) and the x-axis represents the mean of the two methods. The solid red line indicates the mean bias, and the dashed grey lines indicate the 95% limits of agreement (mean bias ± 1.96 SD). Each point represents an individual trial.\\u003c/em\\u003e\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"4.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-9424668/v1/7b34d29b9c0105f61d05a0ed.png\"},{\"id\":107536547,\"identity\":\"03a5c026-51f1-4c4e-a17d-273ced63f07d\",\"added_by\":\"auto\",\"created_at\":\"2026-04-22 11:27:20\",\"extension\":\"png\",\"order_by\":5,\"title\":\"Figure 5\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":93911,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cem\\u003eBland–Altman plots for 6MWT outcomes. Panels show agreement between motion-capture–derived reference and AR-derived measures outcomes for secondary 6MWT metrics (i.e., total steps, number of laps, cadence, and average step length). For each panel, the y-axis represents the difference between methods (Motion capture − AR) and the x-axis represents the mean of the two methods. The solid red line indicates the mean bias, and the dashed grey lines indicate the 95% limits of agreement (mean bias ± 1.96 SD). Each point represents an individual trial.\\u003c/em\\u003e\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"5.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-9424668/v1/ba9a94b76b4b2c0aedef28b7.png\"},{\"id\":107706358,\"identity\":\"b98fd792-99a5-4446-aa6e-96c574e32511\",\"added_by\":\"auto\",\"created_at\":\"2026-04-24 09:17:56\",\"extension\":\"pdf\",\"order_by\":0,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"manuscript-pdf\",\"size\":1022237,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"manuscript.pdf\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-9424668/v1/69722d57-6cc4-4edb-8a60-29d91db389be.pdf\"}],\"financialInterests\":\"Competing interest reported. E.N. is VP of Data and AI, M.R. is a scientific advisor, and E.W. is a software developer for Strolll, a provider of digital neurorehabilitation software via AR glasses. The funder had no role in the design of the study, in the collection, analysis, or interpretation of data, in the writing of the manuscript, or in the decision to publish the results. \\nYR is a member of the Editorial Board of Journal of NeuroEngineering and Rehabilitation. YR was not involved in the journal’s peer review process of, or decisions related to, this manuscript.\",\"formattedTitle\":\"Motor and Balance Assessment in Multiple Sclerosis Using Augmented Reality: Concurrent Validation Against Laboratory Reference Measures\",\"fulltext\":[{\"header\":\"INTRODUCTION\",\"content\":\"\\u003cp\\u003eMultiple sclerosis (MS) is a chronic neurological condition frequently associated with gait, balance, and non-motor impairments that evolve across the disease course and contribute substantially to functional limitations, mobility loss, and reduced quality of life [\\u003cspan additionalcitationids=\\\"CR2 CR3\\\" citationid=\\\"CR1\\\" class=\\\"CitationRef\\\"\\u003e1\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR4\\\" class=\\\"CitationRef\\\"\\u003e4\\u003c/span\\u003e]. Subtle deterioration in postural control and movement coordination can appear from the earliest clinically recognised stages of MS (and often before diagnosis), even when traditional clinical scales (e.g., EDSS, 9HPT, T25FW [\\u003cspan citationid=\\\"CR5\\\" class=\\\"CitationRef\\\"\\u003e5\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR6\\\" class=\\\"CitationRef\\\"\\u003e6\\u003c/span\\u003e]) remain stable [\\u003cspan additionalcitationids=\\\"CR8\\\" citationid=\\\"CR7\\\" class=\\\"CitationRef\\\"\\u003e7\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR9\\\" class=\\\"CitationRef\\\"\\u003e9\\u003c/span\\u003e]. This can be partly explained by the fact that commonly used clinical assessments of motor function in MS are designed primarily for broad disability staging and therefore offer limited granularity, relying on coarse scoring systems (often with significant inter-rater variability) that are not well suited to tracking small but meaningful within-person changes in motor output [\\u003cspan citationid=\\\"CR10\\\" class=\\\"CitationRef\\\"\\u003e10\\u003c/span\\u003e]. These limitations are further compounded by clinical assessments that depend on brief, clinic-based observation, non-standardised procedures, and subjective and categorical scoring [\\u003cspan citationid=\\\"CR5\\\" class=\\\"CitationRef\\\"\\u003e5\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR6\\\" class=\\\"CitationRef\\\"\\u003e6\\u003c/span\\u003e]. Finally, infrequent clinical evaluations can often miss short-lived or fluctuating signs, as they provide sparse sampling of a disease course characterised by day-to-day variability, relapses (especially relevant in the most common form of MS), and treatment-related change; as a result, clinically relevant change may go unrecognised between visits and important therapeutic decisions may be delayed.\\u003c/p\\u003e \\u003cp\\u003eUnlike fully immersive virtual reality, augmented reality (AR) preserves natural vision and real-world perception, making it better suited for balance and gait assessments where safety and more ecologically representative behaviours are essential. Specifically, AR technologies offer an opportunity to overcome the existing limitations of traditional MS scales by enabling standardised, flexible, and highly reproducible motor assessments that can be performed not only in clinical spaces but also in community settings [\\u003cspan citationid=\\\"CR11\\\" class=\\\"CitationRef\\\"\\u003e11\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR12\\\" class=\\\"CitationRef\\\"\\u003e12\\u003c/span\\u003e]. Further, current-generation AR glasses provide multimodal sensing capabilities (e.g. accelerometer, gyroscope, eye tracking), real-time spatial mapping, and controlled visual cues that can be deployed in laboratory as well as home settings for both monitoring and training purposes [\\u003cspan additionalcitationids=\\\"CR13\\\" citationid=\\\"CR12\\\" class=\\\"CitationRef\\\"\\u003e12\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR14\\\" class=\\\"CitationRef\\\"\\u003e14\\u003c/span\\u003e].\\u003c/p\\u003e \\u003cp\\u003eProject PARAMS (Care closer to home: motor-function Parameters from Augmented-Reality-supported Assessments in people living with Multiple Sclerosis) was developed through a partnership between Torbay and South Devon NHS Foundation Trust, Strolll, and the University of Exeter, with the goal of delivering an AR-enabled platform capable of supporting digital assessment of balance and gait, in people with MS. The proposed assessment uses AR glasses (Magic Leap 2) combined with Strolll\\u0026rsquo;s proprietary algorithms to extract spatiotemporal metrics from motor tasks widely used in clinical assessments of posture and gait. However, before Strolll\\u0026rsquo;s software outputs can support clinical monitoring or remote assessment, technical validation against gold-standard instrumentation is required.\\u003c/p\\u003e \\u003cp\\u003eThe tasks included in the PARAMS study were developed with input from a steering group including people with a diagnosis of MS and clinicians. While the main structure for each task was selected based on clinical relevance, details of how the user, clinician and technology interact to produce the necessary calibrations and settings for specific conditions were largely based on recommendations from stakeholders.\\u003c/p\\u003e \\u003cp\\u003eWearable inertial measurement units (IMUs) have emerged as an accessible and relatively low-cost option for capturing gait and balance parameters, and several systems have shown promise for use in MS assessment and other neurodegenerative diseases [\\u003cspan additionalcitationids=\\\"CR16\\\" citationid=\\\"CR15\\\" class=\\\"CitationRef\\\"\\u003e15\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR17\\\" class=\\\"CitationRef\\\"\\u003e17\\u003c/span\\u003e]. However, IMU-based approaches remain constrained by various sources of error, including sensitivity to sensor placement, magnetic disturbance, drift, and reduced precision especially when worn for long durations [\\u003cspan additionalcitationids=\\\"CR19\\\" citationid=\\\"CR18\\\" class=\\\"CitationRef\\\"\\u003e18\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR20\\\" class=\\\"CitationRef\\\"\\u003e20\\u003c/span\\u003e]. AR-based assessments such as those used in PARAMS could eventually help to mitigate these limitations by leveraging integrated eye-tracking and spatial-mapping capabilities. Eye tracking and spatial-mapping provide high-frequency information on visual attention and gaze behaviour that could potentially support more accurate kinematic estimation and reduce reliance on inertial signals alone [\\u003cspan citationid=\\\"CR21\\\" class=\\\"CitationRef\\\"\\u003e21\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR22\\\" class=\\\"CitationRef\\\"\\u003e22\\u003c/span\\u003e]. Eye tracking, specifically, can be useful for verifying adherence to task instructions, for example by confirming whether participants maintain visual fixation or attend to specified visual targets during the assessment. Furthermore, the AR environment enables the safe presentation of interactive digital objects, supporting exergaming-style tasks that enhance engagement without introducing physical hazards for individuals with balance or mobility impairments [\\u003cspan citationid=\\\"CR23\\\" class=\\\"CitationRef\\\"\\u003e23\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR24\\\" class=\\\"CitationRef\\\"\\u003e24\\u003c/span\\u003e]. Together, these features may offer distinct advantages over stand-alone IMU systems by combining richer sensing modalities with standardised and adaptive tasks that are designed to deliver meaningful measurements while prioritising safety, particularly regarding the avoidance of task-irrelevant hazards.\\u003c/p\\u003e \\u003cp\\u003eThe present study aimed to evaluate the concurrent validity of Strolll motor outcomes of widely used clinical assessments of balance and gait against gold-standard motion capture and force plate measurements. We tested the Strolll algorithms across a set of functional tasks commonly used in MS assessment, including standing balance, transitional mobility, and sustained walking. By comparing AR-derived outcomes with reference biomechanical data in both people with MS and healthy controls, this study provides the first comprehensive technical validation of these AR-based motor assessments in this population. Establishing measurement validity and accuracy is an essential step towards future clinical adoption, remote monitoring applications, and integration of AR-based assessments within MS care pathways.\\u003c/p\\u003e\"},{\"header\":\"METHODS\",\"content\":\"\\u003cdiv id=\\\"Sec3\\\" class=\\\"Section2\\\"\\u003e\\n \\u003ch2\\u003eParticipants\\u003c/h2\\u003e\\n \\u003cp\\u003eA total of 46 adults took part in the study, including 26 people with multiple sclerosis (MS) and 20 healthy controls (HC). Participants with MS were recruited through the local community, clinical networks, and social media adverts between March and September 2025. Eligibility criteria for the MS group included a neurologist-confirmed diagnosis, age 18 years or older, and the ability to walk independently for at least one minute with or without a walking aid. Participants were excluded if they had experienced a relapse or major medication change within the previous 30 days, or if they presented with comorbid neurological, psychiatric, or musculoskeletal conditions that could influence balance or gait. Individuals with uncorrected visual impairments or moderate-to-severe cognitive impairment (Mini-Cog\\u0026thinsp;\\u0026lt;\\u0026thinsp;3 [\\u003cspan citationid=\\\"CR25\\\" class=\\\"CitationRef\\\"\\u003e25\\u003c/span\\u003e]) were also excluded. Healthy controls were required to have no history of neurological, vestibular, or mobility-affecting musculoskeletal conditions. Individuals who normally wore glasses for walking were asked to use contact lenses during testing to prevent visual obstruction inside the augmented-reality glasses.\\u003c/p\\u003e\\n \\u003cp\\u003eThe study was approved by the University of Exeter Public Health and Sports Sciences Research Ethics Committee (IRB: 6551385), and all participants provided written informed consent prior to participation.\\u003c/p\\u003e\\n\\u003c/div\\u003e\\n\\u003ch3\\u003eProcedures\\u003c/h3\\u003e\\n\\u003cp\\u003eTesting took place during a single session at the Biomechanics Laboratory at St Luke\\u0026rsquo;s Campus, University of Exeter. Before participant arrival, the laboratory space was scanned using the built-in Magic Leap 2 spatial-mapping routine to allow stable placement of AR objects and reliable operation of its simultaneous localisation and mapping. Participants were then fitted with the augmented-reality glasses, adjusting the rear strap for a secure fit and the compute unit was attached either to their trousers\\u0026apos; waistband or to a belt provided by the experimenters posteriorly at the L5 level.\\u003c/p\\u003e\\n\\u003cp\\u003ePassive reflective markers were placed on anatomical landmarks following the Plug-In-Gait protocol [\\u003cspan citationid=\\\"CR26\\\" class=\\\"CitationRef\\\"\\u003e26\\u003c/span\\u003e], with minor modifications to accommodate wearing AR glasses and its compute unit. Markers were placed on the feet (toe, calcaneus, lateral malleolus), knees (lateral condyle), pelvis (bilateral anterior superior iliac spines and compute pack as a proxy for posterior iliac spines), and on the AR glasses to capture head motions (2 on each side). Participants were familiarised with both the weight of the AR glasses and the augmented-reality environment. The anthropometric calibration procedure was then completed to derive participant-specific parameters such as standing height, seated height, and functional limb lengths.\\u003c/p\\u003e\\n\\u003cp\\u003eParticipants completed three motor tasks presented through the AR interface: a standing balance assessment (Romberg test), the Timed Up and Go (TUG) test, and the Six-Minute Walk Test (6MWT). All tasks were performed while motion-capture cameras and force plates synchronously recorded biomechanical data. Participants were allowed to rest between tasks and a researcher remained alongside them at all times during testing and breaks to ensure safety.\\u003c/p\\u003e\\n\\u003cp\\u003eThe \\u003cstrong\\u003eRomberg task\\u003c/strong\\u003e included eight 30-second trials featuring different combinations of visual condition (eyes open or closed), stance (feet together or tandem), and support surface (firm or foam). Visual cues for the eyes-open conditions were provided by an AR red balloon (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig1\\\" class=\\\"InternalRef\\\"\\u003e1\\u003c/span\\u003e panel A). For tandem stance, the participant\\u0026rsquo;s preferred forward foot was selected during the familiarisation and maintained throughout the rest of the tandem conditions. Participants who used walking aids were permitted to use them during firm-surface trials but were not asked to perform the foam-surface trials. Experimenters ensured that the walking aids were fully on the force plate.\\u003c/p\\u003e\\n\\u003cp\\u003eThe \\u003cstrong\\u003eTUG task\\u003c/strong\\u003e required participants to stand from a chair without arms, walk three metres toward an AR pole, turn around it, and return to sit (Fig. \\u003cspan refid=\\\"Fig1\\\" class=\\\"InternalRef\\\"\\u003e1\\u003c/span\\u003e, panel B). They were instructed that they could walk around the cone in either direction. A verbal \\u0026ldquo;GO\\u0026rdquo; prompt initiated each trial. Both single-task and dual-task conditions were performed. In the dual-task condition, participants completed a verbal Stroop task in which the words \\u0026ldquo;high\\u0026rdquo; or \\u0026ldquo;low\\u0026rdquo; were spoken in either a high or low pitch. Participants were required to repeat the pitch they heard, regardless of word meaning, and were familiarised with this task before performing the dual-task TUG.\\u003c/p\\u003e\\n\\u003cp\\u003eThe \\u003cstrong\\u003e6MWT\\u003c/strong\\u003e required participants to walk continuously for six minutes along a figure-of-eight path defined by two virtual poles placed 3 metres apart (Fig. \\u003cspan refid=\\\"Fig1\\\" class=\\\"InternalRef\\\"\\u003e1\\u003c/span\\u003e, panel C). Participants could use their usual walking aid if required. Participants who wanted to stop before the end of the trial were instructed to stop and stand still, until the experimenter stopped the recording.\\u003c/p\\u003e\\n\\u003ch3\\u003eData Analysis and Outcomes\\u003c/h3\\u003e\\n\\u003cp\\u003eAll laboratory reference data were collected using an 18-camera optical motion-capture system (Vicon, UK, 100 Hz) and, for the Romberg test, strain-gauge force plates (AMTI, 1000 Hz). Motion-capture trajectories and force-plate data were synchronised in Vicon Nexus. Small trajectory gaps shorter than 100 ms were linearly interpolated; longer gaps were reconstructed using rigid-body and kinematic fitting algorithms within Nexus. Kinematic signals were then processed in MATLAB (MathWorks, USA) using the BTK Toolkit functions. Head position and pelvis position were computed as the centroid of four head or pelvis markers. Vertical displacement of heel markers (left and right) was used to detect step events. Marker data were sampled low-pass filtered using a 4th-order zero-lag Butterworth filter (15 Hz, [\\u003cspan citationid=\\\"CR27\\\" class=\\\"CitationRef\\\"\\u003e27\\u003c/span\\u003e]). To allow direct comparison with AR-derived metrics, trial start was defined identically for both systems in each task, ensuring consistent temporal alignment (see each test for further details).\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cspan\\u003e\\u003c/span\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003e1. Romberg\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eCentre-of-pressure (COP) trajectories (mediolateral and anteroposterior) were estimated from ground-reaction forces and moments. COP data were low-pass filtered using a 4th-order, zero-lag Butterworth filter with a 10 Hz cut-off [\\u003cspan citationid=\\\"CR28\\\" class=\\\"CitationRef\\\"\\u003e28\\u003c/span\\u003e]. At the start and end of each trial, the experimenter stepped onto a separate force plate that was hardware-synchronised with the main force-plate system while following the countdown displayed in the AR glasses. Synchronisation with the AR system was performed offline by identifying the two vertical-force peaks recorded by the secondary force plate (threshold: 50 N; minimum temporal distance: 15 s), using the second peak as the alignment event. A 29-s window immediately preceding this event was analysed for the test.\\u003c/p\\u003e\\n\\u003cp\\u003eSway was quantified using the 95% confidence ellipse area [for a definition see [\\u003cspan citationid=\\\"CR29\\\" class=\\\"CitationRef\\\"\\u003e29\\u003c/span\\u003e]]. COP trajectories were mean-centred, the 2\\u0026times;2 covariance matrix computed, and principal axes derived via eigen-decomposition. The ellipse was scaled using the chi-square value for two degrees of freedom (\\u0026chi;\\u0026sup2; = 5.991), and the resulting area (mm\\u0026sup2;) was used as the primary outcome of the test.\\u003c/p\\u003e\\n\\u003cp\\u003eAR sway area estimates were derived from head-motion sensors in the AR glasses. Because these measurements reflect head displacement rather than COP motion measured by the force plates, a calibration step was applied to map AR-derived sway areas into the COP measurement space. Calibration was performed using a log\\u0026ndash;log linear regression across all paired trials (force plate vs. AR raw sway area) using the following formula:\\u003c/p\\u003e\\n\\u003cp\\u003elog10(force plate sway area)\\u0026thinsp;=\\u0026thinsp;0.607 \\u0026times; log10(AR raw sway area)\\u0026thinsp;+\\u0026thinsp;1.168\\u003c/p\\u003e\\n\\u003cp\\u003eThis regression estimated the relationship between the two measurement modalities rather than scaling values to a population average. Because the regression was derived across all trials and produced a near-linear relationship between modalities, the calibration implicitly accounted for differences in sensor location and postural control strategies while preserving trial-to-trial variation in sway magnitude. The resulting transformation was then applied to all AR sway estimates to produce calibrated sway area values used in the final analysis.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cspan\\u003e\\u003c/span\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003e2. Timed Up and Go (TUG)\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThe start of the task was identified using vertical head displacement: a baseline was computed from the first 3 seconds of seated data, and task onset was defined as the first frame at which head height exceeded this baseline by more than 3 SD. Task end was defined as the first frame at which head height returned to baseline within the same threshold for at least 3 consecutive seconds. Automated segmentation was visually inspected and manually corrected when required.\\u003c/p\\u003e\\n\\u003cp\\u003eFor each trial, 4 outcomes were extracted: (1) task duration, defined as the interval between detected start and end frames; (2) total number of steps; (3) mean gait velocity, as the cumulative displacement of the head during the task divided by task duration; (4) total distance. Task duration and total distance were considered primary outcomes, whereas total number of steps and mean gait velocity secondary outcomes.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cspan\\u003e\\u003c/span\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003e3. Six-Minute Walk Test (6MWT)\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThe 6MWT was analysed using head and heel trajectories. Trial start was detected from forward head velocity: walking commenced when absolute head velocity exceeded baseline (first 3 s) by \\u0026gt;\\u0026thinsp;3 SD. The end of the trial was identified when velocity returned below this threshold. Automated segmentation was visually inspected and manually corrected when required.\\u003c/p\\u003e\\n\\u003cp\\u003eSummary outputs for the test included the same 4 outcomes previously described for the TUG test. Additionally, cadence, step length and number of laps were estimated. First, to identify laps, the 2D head trajectory was projected onto its first principal component, and alternating changes in sign of smoothed velocity were used to detect turning points. Consecutive turn pairs defined individual laps. Task duration and total distance were considered primary outcomes, whereas total number of steps, average step length, cadence, number of laps and mean gait velocity secondary outcomes.\\u003c/p\\u003e\\n\\u003ch3\\u003eAR Data Processing\\u003c/h3\\u003e\\n\\u003cp\\u003eData from the augmented-reality glasses were processed using Strolll\\u0026rsquo;s proprietary algorithms embedded within their software code. All outcomes for both AR and gold-standard datasets were exported in standardised formats to ensure reproducibility and comparability across tasks. To support transparency and generalisability, Strolll\\u0026rsquo;s algorithms used for AR-based outcome generation are provided as pseudocode in the study data repository (see Availability of Data and Materials).\\u003c/p\\u003e\\n\\u003cdiv id=\\\"Sec7\\\" class=\\\"Section2\\\"\\u003e\\n \\u003ch2\\u003eStatistical Analysis\\u003c/h2\\u003e\\n \\u003cp\\u003eOutlier handling was task-specific and designed to preserve physiologically plausible variability while excluding confirmed measurement failures. Differences between the gold-standard and AR measurements were converted to Z-scores. Observations exceeding\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;3 standard deviations of the difference distribution were flagged as potential outliers and subsequently reviewed to confirm that they were not attributable to system error (e.g., hardware malfunction, premature trial termination, segmentation failure). Trials affected by confirmed system errors were excluded from analysis. If no device failure was identified, flagged observations were retained in the dataset and analysed.\\u003c/p\\u003e\\n \\u003cp\\u003eConcurrent validity between AR-derived outcomes and gold-standard biomechanical measurements was assessed for each task. Relative agreement between the two measurement systems were examined using intraclass correlation coefficients (ICC two-way random effects, absolute agreement, single measure). ICC values were interpreted according to Koo \\u0026amp; Li[\\u003cspan citationid=\\\"CR30\\\" class=\\\"CitationRef\\\"\\u003e30\\u003c/span\\u003e]: \\u0026lt;0.50 poor, 0.50\\u0026ndash;0.75 moderate, 0.75\\u0026ndash;0.90 good, and \\u0026gt;\\u0026thinsp;0.90 excellent agreement. For 6MWT duration, ICCs were not computed because all participants completed the assessment, resulting in a restricted range and insufficient between-subject variability for a meaningful ICC estimate. Absolute agreement was evaluated using Bland\\u0026ndash;Altman plots, including mean bias and 95% limits of agreement. In these plots, the mean bias (shown as a red line) reflects the average systematic difference between methods, whereas the 95% limits of agreement (shown as dashed lines) indicate the range within which approximately 95% of paired differences are expected to lie. Bias close to zero, narrower limits of agreement, and the absence of an obvious trend across the measurement range were considered indicative of better agreement, whereas a trend in differences with increasing magnitude suggested proportional bias[\\u003cspan citationid=\\\"CR31\\\" class=\\\"CitationRef\\\"\\u003e31\\u003c/span\\u003e]. Mean and standard deviation values for both the gold-standard and AR measures were also reported, together with mean absolute error (MAE) and root mean square error (RMSE). Analyses were conducted for the full sample and separately for the MS and HC groups (except for the Romberg test, due to the limited sample size). Level of significance was set at p\\u0026thinsp;\\u0026lt;\\u0026thinsp;0.05. All analyses were performed in custom scripts written in MATLAB (Mathworks, R2023b).\\u003c/p\\u003e\\n\\u003c/div\\u003e\"},{\"header\":\"Results\",\"content\":\"\\u003cp\\u003eTwenty healthy controls and twenty-six participants with a diagnosis of MS were tested. Three participants with MS and one HC were excluded from the analysis due to technical issues during data collection. People with MS had a mean age of 54.1 (\\u0026plusmn;\\u0026thinsp;9.4) years, while HC had a mean age of 44.7 (\\u0026plusmn;\\u0026thinsp;13.5) years. All people with MS had a diagnosis of Relapsing-Remitting MS with the exception of two that had Primary-Progressive MS and one that had Secondary-Progressive MS. The average number of years since diagnosis was 10.6 (\\u0026plusmn;\\u0026thinsp;6.5). Three people with MS preferred to use walking aids during the walking task. Further, due to balance capacities, only 7 people with MS and 16 HC completed successfully the foam subtasks of the Romberg test.\\u003c/p\\u003e\\n\\u003ch3\\u003eRomberg\\u003c/h3\\u003e\\n\\u003cp\\u003eTwo hundred and thirty-three paired balance trials were included in the analysis. Twenty-seven trials were discarded due to technical issues with the recordings. The most common problems were failure of the AR device, failure of the force-plate, and missing synchronisation. Absolute agreement between calibrated AR sway area and force-plate COP sway area was subtask dependent, ranging from poor (feet together, eyes open on foam condition) to good (tandem, eyes open on firm surface condition). Results divided for sub-conditions are reported in Table\\u0026nbsp;\\u003cspan refid=\\\"Tab1\\\" class=\\\"InternalRef\\\"\\u003e1\\u003c/span\\u003e, Bland-Altman plots of agreement are reported in Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003e. Results stratified for both subtask and group are provided in the supplementary materials (Table S1).\\u003c/p\\u003e \\u003cp\\u003e \\u003cdiv class=\\\"gridtable\\\"\\u003e\\u003ctable float=\\\"Yes\\\" id=\\\"Tab1\\\" border=\\\"1\\\"\\u003e \\u003ccaption language=\\\"En\\\"\\u003e \\u003cdiv class=\\\"CaptionNumber\\\"\\u003eTable 1\\u003c/div\\u003e \\u003cdiv class=\\\"CaptionContent\\\"\\u003e \\u003cp\\u003eValidation results for Romberg test outcomes across the different conditions. The table reports the condition, number of paired observations (force plates vs AR), intraclass correlation coefficient (ICC) with 95% confidence intervals, mean, standard deviation (SD) and agreement metrics (mean absolute error, MAE; root mean square error, RMSE).\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/caption\\u003e \\u003ccolgroup cols=\\\"8\\\"\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c1\\\" colnum=\\\"1\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\".\\\" class=\\\"colspec\\\" colname=\\\"c2\\\" colnum=\\\"2\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c3\\\" colnum=\\\"3\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c4\\\" colnum=\\\"4\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" class=\\\"colspec\\\" colname=\\\"c5\\\" colnum=\\\"5\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c6\\\" colnum=\\\"6\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\".\\\" class=\\\"colspec\\\" colname=\\\"c7\\\" colnum=\\\"7\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\".\\\" class=\\\"colspec\\\" colname=\\\"c8\\\" colnum=\\\"8\\\"\\u003e\\u003c/div\\u003e \\u003cthead\\u003e \\u003ctr\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eCondition\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eNumber of Pairs\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eICC(A,1)\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eICC 95% CI\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003eMean\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;SD (mm\\u003csup\\u003e2\\u003c/sup\\u003e)\\u003c/p\\u003e \\u003cp\\u003e(Force plate)\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003eMean\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;SD (mm\\u003csup\\u003e2\\u003c/sup\\u003e)\\u003c/p\\u003e \\u003cp\\u003e(AR)\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003eMAE\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003eRMSE\\u003c/p\\u003e \\u003c/th\\u003e \\u003c/tr\\u003e \\u003c/thead\\u003e \\u003ctbody\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eFeet Together Floor Eyes Open (mm\\u003c/b\\u003e\\u003csup\\u003e\\u003cb\\u003e2\\u003c/b\\u003e\\u003c/sup\\u003e\\u003cb\\u003e)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e36\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.723\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026minus;\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.185 \\u0026minus;\\u0026thinsp;.756\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e707\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;548\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e680\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;374\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e418\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e650\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eFeet Together Floor Eyes Closed (mm\\u003c/b\\u003e\\u003csup\\u003e\\u003cb\\u003e2\\u003c/b\\u003e\\u003c/sup\\u003e\\u003cb\\u003e)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e36\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.534\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026minus;\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.028 \\u0026minus;\\u0026thinsp;.775\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e1319\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;1299\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e1067\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;951\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e961\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e1775\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eFeet Together Foam Eyes Open (mm\\u003c/b\\u003e\\u003csup\\u003e\\u003cb\\u003e2\\u003c/b\\u003e\\u003c/sup\\u003e\\u003cb\\u003e)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e23\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.488\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026minus;\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e\\u0026minus;\\u0026thinsp;.061 \\u0026minus;\\u0026thinsp;.645\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e900\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;461\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e824\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;583\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e737\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e958\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eFeet Together Foam Eyes Closed (mm\\u003c/b\\u003e\\u003csup\\u003e\\u003cb\\u003e2\\u003c/b\\u003e\\u003c/sup\\u003e\\u003cb\\u003e)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e22\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.443\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026minus;\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.142 \\u0026minus;\\u0026thinsp;.775\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e1713\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;1021\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e1846\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;3899\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e797\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e1084\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eTandem Floor Eyes Open (mm\\u003c/b\\u003e\\u003csup\\u003e\\u003cb\\u003e2\\u003c/b\\u003e\\u003c/sup\\u003e\\u003cb\\u003e)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e35\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.671\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026minus;\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.719 \\u0026minus;\\u0026thinsp;.936\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e896\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;1015\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e1323\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;1008\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e441\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e750\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eTandem Floor Eyes Closed (mm\\u003c/b\\u003e\\u003csup\\u003e\\u003cb\\u003e2\\u003c/b\\u003e\\u003c/sup\\u003e\\u003cb\\u003e)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e35\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.505\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026minus;\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.305 \\u0026minus;\\u0026thinsp;.808\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e8820\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;10441\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e5216\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;4852\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e6273\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e12226\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eTandem Foam Eyes Open (mm\\u003c/b\\u003e\\u003csup\\u003e\\u003cb\\u003e2\\u003c/b\\u003e\\u003c/sup\\u003e\\u003cb\\u003e)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e23\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.936\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026minus;\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.400 \\u0026minus;\\u0026thinsp;.869\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e1353\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;2155\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e1814\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;2553\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e479\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e693\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eTandem Foam Eyes Closed (mm\\u003c/b\\u003e\\u003csup\\u003e\\u003cb\\u003e2\\u003c/b\\u003e\\u003c/sup\\u003e\\u003cb\\u003e)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e23\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.615\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026minus;\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.239 \\u0026minus;\\u0026thinsp;.862\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e9296\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;10175\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e6376\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;5487\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e6817\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e10977\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003c/tbody\\u003e \\u003c/colgroup\\u003e \\u003c/table\\u003e\\u003c/div\\u003e \\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003cp\\u003eFigure \\u003cspan refid=\\\"Fig2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003e \\u003cb\\u003ehere.\\u003c/b\\u003e\\u003c/p\\u003e \\u003cp\\u003eFigure \\u003cspan refid=\\\"Fig2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003e. Bland\\u0026ndash;Altman plot for Romberg sway area. The plot shows agreement between force-plate\\u0026ndash;derived centre-of-pressure (CoP) sway area and the corresponding AR-derived sway area across Romberg conditions. The y-axis represents the difference between methods (Force plates\\u0026thinsp;\\u0026minus;\\u0026thinsp;AR) and the x-axis represents the mean of the two methods. The solid red line indicates the mean bias, and the dashed grey lines indicate the 95% limits of agreement (mean bias\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;1.96 SD). Each point represents an individual trial. Panel A shows the feet together subtasks, while Panel B shows the Tandem subtasks.\\u003c/p\\u003e\\n\\u003ch3\\u003eTimed Up and Go (TUG)\\u003c/h3\\u003e\\n\\u003cp\\u003eA total of 76 matched pairs were included in the final analysis. Six trials were excluded, four due to marker occlusion and two due to AR system failure. Detailed results of the group analysis (HC: 19; MS: 22) are reported in Table\\u0026nbsp;\\u003cspan refid=\\\"Tab2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003e, whereas Bland-Altman plots of agreement are reported in Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig3\\\" class=\\\"InternalRef\\\"\\u003e3\\u003c/span\\u003e.\\u003c/p\\u003e \\u003cp\\u003e \\u003cdiv class=\\\"gridtable\\\"\\u003e\\u003ctable float=\\\"Yes\\\" id=\\\"Tab2\\\" border=\\\"1\\\"\\u003e \\u003ccaption language=\\\"En\\\"\\u003e \\u003cdiv class=\\\"CaptionNumber\\\"\\u003eTable 2\\u003c/div\\u003e \\u003cdiv class=\\\"CaptionContent\\\"\\u003e \\u003cp\\u003eValidation results for Timed Up and Go (TUG) outcomes. Single-task (ST) and dual-task (DT) trials were pooled. For healthy controls (HC) and participants with multiple sclerosis (MS), the table reports agreement for each outcome (duration, mean gait velocity, total distance, and number of steps), quantified using ICC with 95% confidence intervals, mean, standard deviation (SD), mean absolute error (MAE), and root mean square error (RMSE).\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/caption\\u003e \\u003ccolgroup cols=\\\"8\\\"\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c1\\\" colnum=\\\"1\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c2\\\" colnum=\\\"2\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c3\\\" colnum=\\\"3\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c4\\\" colnum=\\\"4\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" class=\\\"colspec\\\" colname=\\\"c5\\\" colnum=\\\"5\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" class=\\\"colspec\\\" colname=\\\"c6\\\" colnum=\\\"6\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c7\\\" colnum=\\\"7\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c8\\\" colnum=\\\"8\\\"\\u003e\\u003c/div\\u003e \\u003cthead\\u003e \\u003ctr\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMetric\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eGroup\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eICC(A,1)\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eICC 95% CI\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003eMean\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;SD\\u003c/p\\u003e \\u003cp\\u003e(Motion Capture)\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003eMean\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;SD\\u003c/p\\u003e \\u003cp\\u003e(AR)\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003eMAE\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003eRMSE\\u003c/p\\u003e \\u003c/th\\u003e \\u003c/tr\\u003e \\u003c/thead\\u003e \\u003ctbody\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\" morerows=\\\"1\\\" rowspan=\\\"2\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eDuration (s)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eHC\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.975\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026minus;\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e[.951 \\u0026minus;\\u0026thinsp;.988]\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e11.3\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;4.0\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e12.8\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;4.4\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e.32\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e.88\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eMS\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.954\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026minus;\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e[.917 \\u0026minus;\\u0026thinsp;.975]\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e15.9\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;7.3\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e17.3\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;7.7\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e.56\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e2.25\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\" morerows=\\\"1\\\" rowspan=\\\"2\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eMean Gait Velocity (m/s)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eHC\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.869\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026minus;\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e[.741 \\u0026minus;\\u0026thinsp;.935]\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e.59 \\u0026plusmn; .07\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e.75 \\u0026plusmn; .15\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e.05\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e.07\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eMS\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.903\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026minus;\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e[.818 \\u0026minus;\\u0026thinsp;.948]\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e.57 \\u0026plusmn; .14\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e.59 \\u0026plusmn; .15\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e.03\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e.06\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\" morerows=\\\"1\\\" rowspan=\\\"2\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eTotal Distance (m)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eHC\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.989\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026minus;\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e[.942 \\u0026minus;\\u0026thinsp;.996]\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e7.9\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;2.0\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e7.7\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;1.9\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e.19\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e.30\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eMS\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.943\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026minus;\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e[.887 \\u0026minus;\\u0026thinsp;.971]\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e8.1\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;1.2\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e7.9\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;1.2\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e.25\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e.40\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\" morerows=\\\"1\\\" rowspan=\\\"2\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eTotal Steps (number)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eHC\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.863\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026minus;\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e[.734 \\u0026minus;\\u0026thinsp;.931]\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e15.2\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;5.2\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e14.2\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;5.3\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e1.74\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e2.81\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eMS\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.701\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026minus;\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e[.443 \\u0026minus;\\u0026thinsp;.841]\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e20.6\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;8.1\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e17.8\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;6.6\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e3.88\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e5.96\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003c/tbody\\u003e \\u003c/colgroup\\u003e \\u003c/table\\u003e\\u003c/div\\u003e \\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003cp\\u003eFigure \\u003cspan refid=\\\"Fig3\\\" class=\\\"InternalRef\\\"\\u003e3\\u003c/span\\u003e \\u003cb\\u003ehere\\u003c/b\\u003e\\u003c/p\\u003e \\u003cp\\u003eFigure \\u003cspan refid=\\\"Fig3\\\" class=\\\"InternalRef\\\"\\u003e3\\u003c/span\\u003e. \\u003cem\\u003eBland\\u0026ndash;Altman plots for Timed Up and Go (TUG) outcomes. Panels show comparisons between motion-capture\\u0026ndash;derived reference outcomes and AR-derived metrics for TUG measures (i.e., duration, mean gait velocity, total distance and step count). For each panel, the y-axis represents the difference between methods (motion capture - AR) and the x-axis the mean of the two methods. The solid red line indicates the mean bias, and the dashed grey lines indicate the 95% limits of agreement (mean bias\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;1.96 SD). Each point represents an individual trial.\\u003c/em\\u003e\\u003c/p\\u003e \\u003cdiv id=\\\"Sec11\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eSix-minute walk test (6MWT)\\u003c/h2\\u003e \\u003cp\\u003eA total of 42 matched pairs were included in the final analysis. Detailed results of the group analysis (HC: 19; MS: 23) are reported in Table\\u0026nbsp;\\u003cspan refid=\\\"Tab3\\\" class=\\\"InternalRef\\\"\\u003e3\\u003c/span\\u003e, whereas Bland-Altman plots of agreements are reported in Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig4\\\" class=\\\"InternalRef\\\"\\u003e4\\u003c/span\\u003e and Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig5\\\" class=\\\"InternalRef\\\"\\u003e5\\u003c/span\\u003e.\\u003c/p\\u003e \\u003cp\\u003e \\u003cdiv class=\\\"gridtable\\\"\\u003e\\u003ctable float=\\\"Yes\\\" id=\\\"Tab3\\\" border=\\\"1\\\"\\u003e \\u003ccaption language=\\\"En\\\"\\u003e \\u003cdiv class=\\\"CaptionNumber\\\"\\u003eTable 3\\u003c/div\\u003e \\u003cdiv class=\\\"CaptionContent\\\"\\u003e \\u003cp\\u003eValidation results for Six-Minute Walk Test (6MWT) outcomes. Results are presented separately for healthy controls (HC) and participants with multiple sclerosis (MS). For each outcome (duration, mean gait velocity, total distance, total steps, number of laps, cadence, and average step length), the table reports agreement quantified using ICC with 95% confidence intervals, mean, standard deviation (SD), mean absolute error (MAE), and root mean square error (RMSE).\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/caption\\u003e \\u003ccolgroup cols=\\\"8\\\"\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c1\\\" colnum=\\\"1\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c2\\\" colnum=\\\"2\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c3\\\" colnum=\\\"3\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c4\\\" colnum=\\\"4\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" class=\\\"colspec\\\" colname=\\\"c5\\\" colnum=\\\"5\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" class=\\\"colspec\\\" colname=\\\"c6\\\" colnum=\\\"6\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\".\\\" class=\\\"colspec\\\" colname=\\\"c7\\\" colnum=\\\"7\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\".\\\" class=\\\"colspec\\\" colname=\\\"c8\\\" colnum=\\\"8\\\"\\u003e\\u003c/div\\u003e \\u003cthead\\u003e \\u003ctr\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMetric\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eGroup\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eICC(A,1)\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eICC 95% CI\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003eMean\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;SD\\u003c/p\\u003e \\u003cp\\u003e(Motion Capture)\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003eMean\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;SD\\u003c/p\\u003e \\u003cp\\u003e(AR)\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003eMAE\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003eRMSE\\u003c/p\\u003e \\u003c/th\\u003e \\u003c/tr\\u003e \\u003c/thead\\u003e \\u003ctbody\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\" morerows=\\\"1\\\" rowspan=\\\"2\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eDuration (s)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eHC\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eNP\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e-\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e360.0 \\u0026plusmn; .2\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e358.9 \\u0026plusmn; .7\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e1.04\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e1.23\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eMS\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eNP\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e-\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e359.9 \\u0026plusmn; .3\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e359.0 \\u0026plusmn; .51\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e.917\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e1.06\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\" morerows=\\\"1\\\" rowspan=\\\"2\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eMean Gait Velocity (m/s)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eHC\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.993\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.980-.997\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e.89 \\u0026plusmn; .11\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e.89 \\u0026plusmn; .12\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e.012\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e.014\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eMS\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.965\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.921 \\u0026minus;\\u0026thinsp;.985\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e.71 \\u0026plusmn; .15\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e.72 \\u0026plusmn; .17\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e.022\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e.044\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\" morerows=\\\"1\\\" rowspan=\\\"2\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eTotal Distance (m)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eHC\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.994\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.983-.998\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e319.8\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;40.0\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e318.9\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;41.0\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e4.03\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e4.77\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eMS\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.964\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.919- .990\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e255.0\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;55.5\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e257.1\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;60.6\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e7.97\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e15.86\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\" morerows=\\\"1\\\" rowspan=\\\"2\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eTotal Steps (number)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eHC\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.741\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.336 \\u0026minus;\\u0026thinsp;.904\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e589.2\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;68.7\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e564.0\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;46.7\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e32.41\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e45.19\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eMS\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.860\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.680 \\u0026minus;\\u0026thinsp;.940\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e566.6\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;80.0\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e550.7\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;59.5\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e29.57\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e38.54\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\" morerows=\\\"1\\\" rowspan=\\\"2\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eNumber of laps (number)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eHC\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.881\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.707 \\u0026minus;\\u0026thinsp;.955\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e33.5\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;4.9\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e32.7\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;5.6\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e1.47\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e2.65\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eMS\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.815\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.617 - .917\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e25.2\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;7.2\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e23.8\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;9.3\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e2.74\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e5.17\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\" morerows=\\\"1\\\" rowspan=\\\"2\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eCadence (steps/min)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eHC\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.742\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.343 \\u0026minus;\\u0026thinsp;.904\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e98.2\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;11.5\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e94.1\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;7.8\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e5.39\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e7.52\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eMS\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.861\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.688 \\u0026minus;\\u0026thinsp;.940\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e94.4\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;13.3\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e91.9\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;10.0\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e4.93\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e6.40\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\" morerows=\\\"1\\\" rowspan=\\\"2\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eAverage Step Length (m)\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eHC\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.580\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.117 \\u0026minus;\\u0026thinsp;.829\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e.54 \\u0026plusmn; .04\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e.57 \\u0026plusmn; .05\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e.027\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e.041\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eMS\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e.900\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e.656 \\u0026minus;\\u0026thinsp;.963\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e.45 \\u0026plusmn; .07\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\"\\u0026plusmn;\\\" colname=\\\"c6\\\"\\u003e \\u003cp\\u003e.47 \\u0026plusmn; .07\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c7\\\"\\u003e \\u003cp\\u003e.025\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c8\\\"\\u003e \\u003cp\\u003e.034\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003c/tbody\\u003e \\u003c/colgroup\\u003e \\u003c/table\\u003e\\u003c/div\\u003e \\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec12\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eFigures \\u003cspan refid=\\\"Fig4\\\" class=\\\"InternalRef\\\"\\u003e4\\u003c/span\\u003e and \\u003cspan refid=\\\"Fig5\\\" class=\\\"InternalRef\\\"\\u003e5\\u003c/span\\u003e Here\\u003c/h2\\u003e \\u003cp\\u003eFigure \\u003cspan refid=\\\"Fig4\\\" class=\\\"InternalRef\\\"\\u003e4\\u003c/span\\u003e. \\u003cem\\u003eBland\\u0026ndash;Altman plots for 6MWT outcomes. Panels show agreement between motion-capture\\u0026ndash;derived reference and AR-derived measures outcomes for primary 6MWT metrics (i.e., duration and distance). For each panel, the y-axis represents the difference between methods (Motion capture\\u0026thinsp;\\u0026minus;\\u0026thinsp;AR) and the x-axis represents the mean of the two methods. The solid red line indicates the mean bias, and the dashed grey lines indicate the 95% limits of agreement (mean bias\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;1.96 SD). Each point represents an individual trial.\\u003c/em\\u003e\\u003c/p\\u003e \\u003cp\\u003eFigure \\u003cspan refid=\\\"Fig5\\\" class=\\\"InternalRef\\\"\\u003e5\\u003c/span\\u003e. \\u003cem\\u003eBland\\u0026ndash;Altman plots for 6MWT outcomes. Panels show agreement between motion-capture\\u0026ndash;derived reference and AR-derived measures outcomes for secondary 6MWT metrics (i.e., total steps, number of laps, cadence, and average step length). For each panel, the y-axis represents the difference between methods (Motion capture\\u0026thinsp;\\u0026minus;\\u0026thinsp;AR) and the x-axis represents the mean of the two methods. The solid red line indicates the mean bias, and the dashed grey lines indicate the 95% limits of agreement (mean bias\\u0026thinsp;\\u0026plusmn;\\u0026thinsp;1.96 SD). Each point represents an individual trial.\\u003c/em\\u003e\\u003c/p\\u003e \\u003c/div\\u003e\"},{\"header\":\"Discussion\",\"content\":\"\\u003cp\\u003eThis study evaluated the concurrent validity of motor and balance outcomes derived from AR glasses, and analysed through Strolll proprietary algorithms, against gold-standard laboratory instrumentation (i.e., motion capture and force plates) in people with MS and age-matched HC. Overall, the AR derived metrics demonstrated task-dependent validity: agreement was poor-to-excellent for static balance (Romberg sway area), and moderate-to-excellent for key functional mobility and walking outcomes (Timed Up and Go; Six-Minute Walk Test). Particularly, agreement was generally strongest for global spatiotemporal parameters (e.g., task duration, mean gait velocity, total distance), whereas step-derived outcomes (e.g., step count, cadence, step length) showed more variable performance. Importantly, the strongest-performing outcomes in this validation correspond to the same primary metrics commonly obtained \\u0026ldquo;by hand\\u0026rdquo; by clinicians in routine clinical practice (e.g., TUG duration, 6MWT total distance). The good-to-excellent agreement observed for these global mobility metrics indicates that the AR-derived estimates provide quantitatively comparable values suitable for assessment and longitudinal monitoring and for early and subtle neurological changes.\\u003c/p\\u003e \\u003cp\\u003eOverall, the pattern observed here - high agreement for global mobility outcomes and more variable agreement for step-level metrics - is consistent with broader validation work in AR and wearable mobility assessment [\\u003cspan citationid=\\\"CR32\\\" class=\\\"CitationRef\\\"\\u003e32\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR33\\\" class=\\\"CitationRef\\\"\\u003e33\\u003c/span\\u003e]. For example, prior validation studies have reported excellent agreement for TUG completion duration and temporal gait outcomes, but weaker performance for some turning- or spatial measures of gait [\\u003cspan citationid=\\\"CR34\\\" class=\\\"CitationRef\\\"\\u003e34\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR35\\\" class=\\\"CitationRef\\\"\\u003e35\\u003c/span\\u003e]. This aligns with the idea that event detection (e.g., discrete step events) and derived measures that depend on accurate segmentation may be more sensitive to algorithm selection and signal quality than continuous measures, and that wearable sensors tend to be less performative for spatial outcomes [\\u003cspan citationid=\\\"CR36\\\" class=\\\"CitationRef\\\"\\u003e36\\u003c/span\\u003e]. The comparatively lower agreement for step-derived outcomes in parts of our dataset likely reflects several interacting factors: first, step detection definitions differ by system (heel-marker events in motion capture vs AR-based estimation), which can introduce small discrepancies that accumulate over time (especially in long tasks such as the 6MWT); second, ICC is sensitive to between-subject variability: when a cohort is relatively homogeneous on a given metric, ICC coefficients can appear lower even when absolute errors are small. This is consistent with our observation that some HC step-derived ICCs were modest despite relatively small MAE/RMSE (see Table\\u0026nbsp;\\u003cspan refid=\\\"Tab1\\\" class=\\\"InternalRef\\\"\\u003e1\\u003c/span\\u003e\\u0026ndash;\\u003cspan refid=\\\"Tab3\\\" class=\\\"InternalRef\\\"\\u003e3\\u003c/span\\u003e). In the following sections, we discuss these task-specific findings in more detail (Romberg, TUG, and 6MWT), and consider their implications for objective monitoring and technology-enabled rehabilitation in MS.\\u003c/p\\u003e \\u003cdiv id=\\\"Sec14\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eRomberg\\u003c/h2\\u003e \\u003cp\\u003eFor quiet-stance balance (i.e., Romberg), calibrated AR sway area demonstrated poor-to-good agreement with force-plate CoP sway area. Agreement varied by condition, ranging from lower values in some feet-together foam conditions to high agreement in tandem stance on firm ground (see Table\\u0026nbsp;\\u003cspan refid=\\\"Tab1\\\" class=\\\"InternalRef\\\"\\u003e1\\u003c/span\\u003e). We interpret these differences as arising from three main factors. First, postural control strategies differ substantially across stance configurations, particularly in tandem stance. Compared with feet-together standing - where sway is often dominated by an ankle strategy and primarily expressed in the anteroposterior direction - tandem stance constrains the base of support and increases the challenge to mediolateral stability [\\u003cspan citationid=\\\"CR37\\\" class=\\\"CitationRef\\\"\\u003e37\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR38\\\" class=\\\"CitationRef\\\"\\u003e38\\u003c/span\\u003e]. Participants may adopt a stiffer control strategy (e.g., increased co-contraction and greater reliance on hip/trunk contributions), with larger mediolateral excursions and a higher likelihood of \\u0026ldquo;side loss of balance\\u0026rdquo;. These strategy changes may increase the coupling between global body motion and head kinematics, thereby improving alignment between AR head-based sway estimates and CoP-derived outcomes in some tandem conditions. Second, Romberg conditions performed on foam introduce a compliant interface between the feet and force plate. This manipulation both alters sensory weighting (e.g., reduced reliable somatosensory/proprioceptive input) and changes the mechanics at the foot\\u0026ndash;support interface, typically resulting in increased CoP excursion and variability. In addition, foam can introduce components in the ground reaction force that reflect surface deformation dynamics (e.g., compression\\u0026ndash;rebound behaviour and shear effects) rather than postural control alone. These factors weaken the direct relationship between CoP and centre-of-mass (CoM) motion assumed by inverted-pendulum models [\\u003cspan citationid=\\\"CR37\\\" class=\\\"CitationRef\\\"\\u003e37\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR38\\\" class=\\\"CitationRef\\\"\\u003e38\\u003c/span\\u003e], and may therefore reduce correspondence between CoP-based metrics and head-based kinematic estimates. Third, the two outcomes are related but not identical: force-plate-based posturography quantifies control at the base of support (i.e., CoP), whereas AR estimates here are derived from head-based sensing. This distinction was intentional. While lumbar-mounted inertial measurement units are widely used to estimate balance in both research and clinical contexts [\\u003cspan citationid=\\\"CR36\\\" class=\\\"CitationRef\\\"\\u003e36\\u003c/span\\u003e], we specifically evaluated whether head-based sensing could provide valid balance-related information with minimal instrumentation. Further, head kinematics may capture functionally relevant aspects of stability - such as upper-body control and head stabilisation required to maintain gaze and manage vestibular demands - which are important in daily activities and may relate to fall risk [\\u003cspan citationid=\\\"CR39\\\" class=\\\"CitationRef\\\"\\u003e39\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR40\\\" class=\\\"CitationRef\\\"\\u003e40\\u003c/span\\u003e]. Lastly, head-based sensing removes the need for floor-mounted measurement devices. Although low-cost systems can provide meaningful balance metrics under standard conditions, their performance can be compromised on compliant surfaces such as the foam used in the Romberg sub-tests. In this respect, AR-based approaches offer a practical advantage [\\u003cspan citationid=\\\"CR41\\\" class=\\\"CitationRef\\\"\\u003e41\\u003c/span\\u003e]. Accordingly, we believe that head-based metrics should not replace force-plate posturography/lumbar-based measures, but to provide a complementary perspective - particularly relevant for scalable monitoring and tasks in which upper-body control is a major contributor to functional stability. Importantly, given the cost, set-up time, and limited portability of force-plate systems, head-based measures collected through portable devices, such as the one employed in this study, may also serve as a minimally instrumented option in contexts where laboratory instrumentation is impractical or unavailable, including remote and field-deployable assessments.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec15\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eTimed Up and Go\\u003c/h2\\u003e \\u003cp\\u003eFor the TUG test, agreement between the AR-based and the motion capture outcomes was consistently strong across outcomes (both primary and secondary) and groups. Specifically, agreement was excellent for duration and total distance, with good agreement for mean gait velocity and step count (see Table\\u0026nbsp;\\u003cspan refid=\\\"Tab2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003e). The strong agreement for duration is consistent with our processing approach, which defines task onset/offset from vertical head displacement (sit-to-stand and return-to-sit). These findings align with prior instrumented-TUG validation work showing that total time can be captured with very high agreement relative to motion capture systems, while reducing reliance on operator-dependent errors (e.g., stopwatch) [\\u003cspan citationid=\\\"CR42\\\" class=\\\"CitationRef\\\"\\u003e42\\u003c/span\\u003e]. From a broader rehabilitation and monitoring perspective, the value of instrumenting the TUG is not simply replicating total time but enabling objective quantification of components that reflect balance control and fall risk (e.g., turning and chair transfers). Nevertheless, demonstrating that automated AR-derived duration closely matches laboratory-derived duration supports immediate applicability as a minimally instrumented alternative to stopwatch timing, while enabling richer and more objective subcomponent metrics in future work. Previous work highlighted that conventional TUG scoring can miss intervention effects because it collapses multiple mobility skills into a single time outcome, whereas instrumented approaches enable objective subcomponent analysis (sit-to-stand, gait, turning, turn-to-sit) and richer motor evaluations [\\u003cspan citationid=\\\"CR43\\\" class=\\\"CitationRef\\\"\\u003e43\\u003c/span\\u003e]. In fact, recent work using IMUs in TUG tasks highlights that turning and sitting-related parameters (e.g., reduced turning angular velocity; altered acceleration during stand-to-sit) can be particularly informative for detecting deterioration in physical function, reinforcing the clinical relevance of subphase and metrics beyond total time [\\u003cspan additionalcitationids=\\\"CR45 CR46\\\" citationid=\\\"CR44\\\" class=\\\"CitationRef\\\"\\u003e44\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR47\\\" class=\\\"CitationRef\\\"\\u003e47\\u003c/span\\u003e]. Further, incorporating a cognitive dual-task during the TUG (e.g., as in the Mini-BESTest [\\u003cspan citationid=\\\"CR48\\\" class=\\\"CitationRef\\\"\\u003e48\\u003c/span\\u003e]) provides an opportunity to investigate cognitive\\u0026ndash;motor interactions, and laboratory measures under dual-task conditions may better reflect everyday mobility demands than single-task assessments [\\u003cspan citationid=\\\"CR49\\\" class=\\\"CitationRef\\\"\\u003e49\\u003c/span\\u003e]. AR-based systems have already shown that several TUG sub-durations can be derived with excellent between-systems agreement, although turning segmentation can be more challenging than straight walking or chair transitions [\\u003cspan citationid=\\\"CR50\\\" class=\\\"CitationRef\\\"\\u003e50\\u003c/span\\u003e]. Taken together, these elements position the present findings as an important enabling step: by demonstrating strong validity for global TUG outcomes, the AR approach is well placed to extend toward richer, phase-specific digital biomarkers in future iterations - particularly relevant for technology-supported rehabilitation pathways that aim to scale objective assessment into routine care and home settings.\\u003c/p\\u003e \\u003cp\\u003eThe comparatively lower agreement for step count in MS likely reflects a combination of algorithmic and behavioural factors. First, step detection is inherently sensitive to atypical movement patterns; as mobility becomes more pathological (e.g., slower speed, altered rhythm, disrupted turning), segmentation and threshold-based detection become more challenging using data from wearable devices and biomechanical laboratory reference systems alike. Second, our protocol allowed testing with or without a walking aid, and assistive device use alters the upper-body motion and intermittently constrains natural gait dynamics, which may affect the accuracy of acceleration-derived step features, compared with spatial reference systems such as motion capture or visual SLAM [\\u003cspan citationid=\\\"CR51\\\" class=\\\"CitationRef\\\"\\u003e51\\u003c/span\\u003e]. Notably, some instrumented-TUG studies exclude participants who use walking aids, which may strengthen statistical accuracy at the cost of clinical generalisability; in contrast, our approach intentionally prioritises inclusion of a wider and more representative clinical cohort, with the trade-off that step-based metrics in walking-aid users should currently be interpreted more cautiously. Future development should therefore focus on improving robustness of step-event detection in the presence of assistive devices and more impaired gait patterns (e.g., adaptive thresholds and/or multi-feature fusion), to better support deployment in heterogeneous rehabilitation cohorts.\\u003c/p\\u003e \\u003cp\\u003e \\u003cb\\u003e6-Minute Walking Test\\u003c/b\\u003e \\u003c/p\\u003e \\u003cp\\u003eFor the Six-Minute Walk Test (6MWT), agreement was excellent for mean gait velocity and total distance and moderate-to-good for laps and step-derived measures (see Table\\u0026nbsp;\\u003cspan refid=\\\"Tab3\\\" class=\\\"InternalRef\\\"\\u003e3\\u003c/span\\u003e). We believe that these results contribute to the existing literature, providing a valid, accurate alternative to provide quantitative assessments of this common test. A key contribution stems from the fact that here the 6MWT was implemented as a figure-of-eight trajectory between two AR poles, rather than as a straight corridor walk. This design choice was pragmatic and clinically motivated as a figure-of-eight walking trajectory demands an equal number of turns in the left and right directions, requires substantially less physical space - a common constraint in clinical environments - and reduces the need for participants to navigate busy corridors, thereby limiting external distractions and making it easier for staff to provide close supervision or physical assistance when needed. Further, this layout is also advantageous for potential home-based assessments, where available walkway length is often limited to a few metres. In addition, a turning-rich trajectory may be particularly valuable from a biomechanical perspective because it increases task complexity relative to straight walking, can help reveal walking asymmetries, and better reflects the demands of everyday mobility where turning is frequent and often challenging. These considerations align with the work of Shah et al., who highlight walkway length and turning content as key determinants of 6MWT distance-estimation error, when using wearable devices, noting that turning is likely a principal contributor to measurement discrepancies when estimating distance from wearable sensors [\\u003cspan citationid=\\\"CR50\\\" class=\\\"CitationRef\\\"\\u003e50\\u003c/span\\u003e]. Despite the high turning content in our protocol (poles separated by 3 m and most times\\u0026thinsp;\\u0026gt;\\u0026thinsp;25 laps completed per participant), AR-derived distance and mean gait velocity demonstrated excellent agreement with motion capture in both HC and MS, suggesting that the AR system can provide robust estimates even under turning-intensive walking conditions. It is however worth mentioning that, similarly to the TUG task, the residual differences observed in a small subset of participants (see Figs.\\u0026nbsp;\\u003cspan refid=\\\"Fig4\\\" class=\\\"InternalRef\\\"\\u003e4\\u003c/span\\u003e and \\u003cspan refid=\\\"Fig5\\\" class=\\\"InternalRef\\\"\\u003e5\\u003c/span\\u003e) may partly reflect the use of walking aids (i.e., walkers and walking sticks). Consequently, we believe that 6MWT AR-derived outcomes in cohorts using walking aids should be interpreted with caution and future developments should explicitly address aid-related movement patterns to improve accuracy and generalisability.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec16\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eTechnology-enabled MS disease monitoring and rehabilitation\\u003c/h2\\u003e \\u003cp\\u003eConventional MS motor assessments can be coarse-grained, partly subjective, and infrequent, which limits sensitivity to subtle within-person change and constrains timely clinical decision-making, especially in the remitting-and-relapse MS subtype. The broader vision of project PARAMS is to enable standardised, objective, home-based motor-function assessment using wearable AR glasses, supporting longitudinal monitoring and trend analysis, and ultimately helping services deliver care \\u0026ldquo;closer to home\\u0026rdquo; while improving efficiency and cost-effectiveness. Within that working frame, the present technical validation provides a critical prerequisite as it demonstrates that several clinically interpretable outcomes - particularly duration, mean gait velocity, and distance across common clinical tests - can be captured with high agreement relative to laboratory gold standards. These outcomes are directly relevant to clinical monitoring and rehabilitation because they are commonly used for baseline profiling, goal setting, and tracking response to therapy or disease progression. In addition, this AR platform can support standardisation (consistent task setup and cueing) and may reduce measurement variability introduced by rater-dependent scoring.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec17\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eLimitations\\u003c/h2\\u003e \\u003cp\\u003eThe study has five main limitations. First, data were collected in a laboratory context, and performance may differ in true home environments (e.g., lighting conditions, space constraints, connectivity issues). Second, while the present work focuses on concurrent validity, the broader implementation case for pathway change depends on additional evidence (feasibility, usability, reliability, and responsiveness to change). Ongoing studies are assessing these domains. Third, our MS cohort was predominantly relapsing\\u0026ndash;remitting, with only a small number of participants with progressive phenotypes. As gait and balance impairment are often greater in progressive MS, the validity of AR-derived outcomes in more advanced or progressive disease (and across wider disability levels) requires confirmation in larger, more phenotypically diverse samples. Fourth, the presence of walking aids during the assessment may reduce the overall accuracy of the AR-based metrics, especially those related to step characteristics, therefore future studies should specifically focus on further developing these algorithms to further expand the generalisability of this approach. Finally, because many participants with MS could not complete the foam subtasks of the Romberg test, we were unable to stratify the validation by group for these conditions.\\u003c/p\\u003e \\u003c/div\\u003e\"},{\"header\":\"Conclusion\",\"content\":\"\\u003cp\\u003eThis study demonstrates that motor outcomes derived from AR glasses and processed with Strolll\\u0026rsquo;s algorithms show task-dependent concurrent validity against laboratory gold standards (motion capture and force plates) in people with MS and age-matched controls. Agreement was strongest for global mobility metrics that are central to routine clinical assessment - particularly TUG duration and distance and 6MWT distance and mean gait velocity - supporting AR-based measurement as a minimally instrumented alternative when laboratory systems are impractical. Step-derived outcomes showed more variable performance, especially in MS, indicating a priority area for algorithm refinement and robustness to impaired gait patterns and walking-aid use. Collectively, these results support AR-based assessment as a valid and scalable approach for objective motor measurement in neurorehabilitation, and performance in real-world clinical and community settings.\\u003c/p\\u003e\"},{\"header\":\"Declarations\",\"content\":\"\\u003cp\\u003e\\u003cstrong\\u003eEthics approval and consent to participate\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThe study was approved by the University of Exeter Public Health and Sports Sciences Research Ethics Committee (IRB: 6551385), and all participants provided written informed consent prior to participation.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eConsent for publication\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eNot applicable.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eAvailability of data and materials\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eTo ensure methodological transparency and reproducibility, motion capture and force plates data as well as all custom MATLAB scripts/functions used for processing and analysing the force-plate and motion-capture datasets (i.e., data import, filtering, segmentation, COP computation, step detection, and gait metric extraction) are available in the Open Science Framework repository: \\u003cu\\u003ehttps://osf.io/fup6j\\u003c/u\\u003e. These shared scripts/functions exclude the proprietary Strolll algorithms used for AR-based outcome generation, which are however shared in the form of pseudocode to ensure proper generalization, and therein negating a \\u0026lsquo;black-box\\u0026rsquo; approach, without sharing proprietary code, in the same repository.\\u003c/p\\u003e\\n\\u003cp\\u003eFor the purpose of open access, the author has applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising from this submission.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eCompeting interests\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eE.N. is VP of Data and AI, M.R. is a scientific advisor, and E.W. is a software developer for Strolll, a provider of digital neurorehabilitation software via AR glasses. The funder had no role in the design of the study, in the collection, analysis, or interpretation of data, in the writing of the manuscript, or in the decision to publish the results.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eYR is a member of the Editorial Board of Journal of NeuroEngineering and Rehabilitation. YR was not involved in the journal\\u0026rsquo;s peer review process of, or decisions related to, this manuscript.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eFunding\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThis study was funded by Torbay and South Devon NHS Foundation Trust via an award from NHS England through the NIHR Digital Health Partnership Award (Project PARAMS). and was supported by the National Institute for Health and Care Research (NIHR) Exeter Biomedical Research Centre. The views expressed are those of the authors and do not necessarily reflect those of NHS England, the NIHR, or the Department of Health and Social Care.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eAuthors\\u0026apos; contributions\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eStudy Conception and design \\u0026ndash; YR, EN, AS, MR \\u0026amp; WY. Acquisition of data \\u0026ndash; YR \\u0026amp; WY. Data curation \\u0026ndash; YR, JY \\u0026amp; PL. Algorithm Development \\u0026ndash; EN \\u0026amp; EW. Study Analysis Software \\u0026ndash; YR \\u0026amp; EN. Formal analysis \\u0026ndash; YR. Visualisation \\u0026ndash; YR \\u0026amp; EN. Funding \\u0026ndash; YR, AS \\u0026amp; WY. Writing original draft - YR \\u0026amp; WY. Writing review \\u0026amp; editing - All the authors. Final approval of the completed article - All the authors.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eAcknowledgements\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eWe would like to thank the study participants and those who accompanied them to the laboratory. We also thank people with MS, the MS coordinator, the MS nurse, and the Torbay MS Research Team who attended the PPIE workshops and provided guidance on the development of the software. \\u0026nbsp;We are grateful to Prof Stefano Pluchino (University of Cambridge) for study design advice and Dr Antonio Scalfari (Imperial College London) for input on the funding application. Finally, we thank Barri Morgan (Strolll) for facilitating communication between the parties involved in PARAMS.\\u003c/p\\u003e\"},{\"header\":\"References\",\"content\":\"\\u003col\\u003e\\n \\u003cli\\u003eReich DS, Lucchinetti CF, Calabresi PA. Multiple Sclerosis. New England Journal of Medicine. Massachusetts Medical Society; 2018;378:169\\u0026ndash;80. https://doi.org/10.1056/nejmra1401483\\u003c/li\\u003e\\n \\u003cli\\u003eComber L, Galvin R, Coote S. Gait deficits in people with multiple sclerosis: A systematic review and meta-analysis. Gait Posture. Elsevier B.V.; 2017. p. 25\\u0026ndash;35. https://doi.org/10.1016/j.gaitpost.2016.09.026\\u003c/li\\u003e\\n \\u003cli\\u003eComber L, Sosnoff JJ, Galvin R, Coote S. Postural control deficits in people with Multiple Sclerosis: A systematic review and meta-analysis. Gait Posture. Elsevier B.V.; 2018. p. 445\\u0026ndash;52. https://doi.org/10.1016/j.gaitpost.2018.02.018\\u003c/li\\u003e\\n \\u003cli\\u003eLane J, Ng HS, Poyser C, Lucas RM, Tremlett H. Multiple sclerosis incidence: A systematic review of change over time by geographical region. Mult. Scler. Relat. Disord. Elsevier B.V.; 2022. https://doi.org/10.1016/j.msard.2022.103932\\u003c/li\\u003e\\n \\u003cli\\u003eUitdehaag BMJ. Clinical outcome measures in multiple sclerosis. In: Goodin D, editor. handbook of clinical neurology, Multiple Sclerosis and Related Disorders. 3rd ed. Elsevier; 2014.\\u003c/li\\u003e\\n \\u003cli\\u003eInojosa H, Schriefer D, Ziemssen T. Clinical outcome measures in multiple sclerosis: A review. Autoimmun. Rev. Elsevier B.V.; 2020. https://doi.org/10.1016/j.autrev.2020.102512\\u003c/li\\u003e\\n \\u003cli\\u003eMarrie RA, Allegretta M, Barcellos LF, Bebo B, Calabresi PA, Correale J, et al. From the prodromal stage of multiple sclerosis to disease prevention. Nat Rev Neurol. Nature Research; 2022;18:559\\u0026ndash;72. https://doi.org/10.1038/s41582-022-00686-x\\u003c/li\\u003e\\n \\u003cli\\u003eMelillo F, Di Sapio A, Martire S, Malentacchi M, Matta M, Bertolotto A. Computerized posturography is more sensitive than clinical Romberg Test in detecting postural control impairment in minimally impaired Multiple Sclerosis patients. Mult Scler Relat Disord. Elsevier B.V.; 2017;14:51\\u0026ndash;5. https://doi.org/10.1016/j.msard.2017.03.008\\u003c/li\\u003e\\n \\u003cli\\u003eVienne-Jumeau A, Quijoux F, Vidal PP, Ricard D. Wearable inertial sensors provide reliable biomarkers of disease severity in multiple sclerosis: A systematic review and meta-analysis. Ann. Phys. Rehabil. Med. Elsevier Masson SAS; 2020. p. 138\\u0026ndash;47. https://doi.org/10.1016/j.rehab.2019.07.004\\u003c/li\\u003e\\n \\u003cli\\u003eInojosa H, Schriefer D, Kl\\u0026ouml;ditz A, Trentzsch K, Ziemssen T. Balance Testing in Multiple Sclerosis\\u0026mdash;Improving Neurological Assessment With Static Posturography? Front Neurol. Frontiers Media S.A.; 2020;11. https://doi.org/10.3389/fneur.2020.00135\\u003c/li\\u003e\\n \\u003cli\\u003eFern\\u0026aacute;ndez-Ca\\u0026ntilde;as M, Cano-de-la-Cuerda R, Marcos-Ant\\u0026oacute;n S, Onate-Figu\\u0026eacute;rez A. Augmented and Mixed Reality Interventions in People with Multiple Sclerosis: A Systematic Review. Brain Sci. Multidisciplinary Digital Publishing Institute (MDPI); 2025. https://doi.org/10.3390/brainsci15121292\\u003c/li\\u003e\\n \\u003cli\\u003evan Bergem JS, van Doorn PF, Hoogendoorn EM, Geerse DJ, Roerdink M. Gait and Balance Assessments with Augmented Reality Glasses in People with Parkinson\\u0026rsquo;s Disease: Concurrent Validity and Test\\u0026ndash;Retest Reliability. Sensors. Multidisciplinary Digital Publishing Institute (MDPI); 2024;24. https://doi.org/10.3390/s24175485\\u003c/li\\u003e\\n \\u003cli\\u003evan Doorn PF, Nyman E, Wishaupt K, van der Krogt MM, Roerdink M. Deriving Motor States and Mobility Metrics from Gamified Augmented Reality Rehabilitation Exercises in People with Parkinson\\u0026rsquo;s Disease. Sensors. Multidisciplinary Digital Publishing Institute (MDPI); 2025;25. https://doi.org/10.3390/s25237172\\u003c/li\\u003e\\n \\u003cli\\u003ePalumbo A. Microsoft HoloLens 2 in Medical and Healthcare Context: State of the Art and Future Prospects. Sensors. MDPI; 2022. https://doi.org/10.3390/s22207709\\u003c/li\\u003e\\n \\u003cli\\u003eRusso Y, Stuart S, Silva-Batista C, Brumbach B, Vannozzi G, Mancini M. Does visual cueing improve gait initiation in people with Parkinson\\u0026rsquo;s disease? Hum Mov Sci. Elsevier B.V.; 2022;84. https://doi.org/10.1016/j.humov.2022.102970\\u003c/li\\u003e\\n \\u003cli\\u003eVoisard C, de l\\u0026rsquo;Escalopier N, Vienne-Jumeau A, Moreau A, Quijoux F, Bompaire F, et al. Innovative multidimensional gait evaluation using IMU in multiple sclerosis: introducing the semiogram. Front Neurol. Frontiers Media SA; 2023;14. https://doi.org/10.3389/fneur.2023.1237162\\u003c/li\\u003e\\n \\u003cli\\u003eBrognara L, Palumbo P, Grimm B, Palmerini L. Assessing Gait in Parkinson\\u0026rsquo;s Disease Using Wearable Motion Sensors: A Systematic Review. Diseases. MDPI AG; 2019;7:18. https://doi.org/10.3390/diseases7010018\\u003c/li\\u003e\\n \\u003cli\\u003eRast FM, Labruy\\u0026egrave;re R. Systematic review on the application of wearable inertial sensors to quantify everyday life motor activity in people with mobility impairments. J. Neuroeng. Rehabil. BioMed Central Ltd; 2020. https://doi.org/10.1186/s12984-020-00779-y\\u003c/li\\u003e\\n \\u003cli\\u003eRobert-Lachaine X, Mecheri H, Larue C, Plamondon A. Effect of local magnetic field disturbances on inertial measurement units accuracy. Appl Ergon. Elsevier Ltd; 2017;63:123\\u0026ndash;32. https://doi.org/10.1016/j.apergo.2017.04.011\\u003c/li\\u003e\\n \\u003cli\\u003eGu C, Lin W, He X, Zhang L, Zhang M. IMU-based motion capture system for rehabilitation applications: A systematic review. Biomimetic Intelligence and Robotics. Elsevier B.V.; 2023. https://doi.org/10.1016/j.birob.2023.100097\\u003c/li\\u003e\\n \\u003cli\\u003eH\\u0026uuml;bner P, Clintworth K, Liu Q, Weinmann M, Wursthorn S. Evaluation of hololens tracking and depth sensing for indoor mapping applications. Sensors (Switzerland). MDPI AG; 2020;20. https://doi.org/10.3390/s20041021\\u003c/li\\u003e\\n \\u003cli\\u003eSoares I, Sousa RB, Petry M, Moreira AP. Accuracy and repeatability tests on hololens 2 and htc vive. Multimodal Technologies and Interaction. MDPI; 2021;5. https://doi.org/10.3390/mti5080047\\u003c/li\\u003e\\n \\u003cli\\u003eCalafiore D, Invernizzi M, Ammendolia A, Marotta N, Fortunato F, Paolucci T, et al. Efficacy of Virtual Reality and Exergaming in Improving Balance in Patients With Multiple Sclerosis: A Systematic Review and Meta-Analysis. Front. Neurol. Frontiers Media S.A.; 2021. https://doi.org/10.3389/fneur.2021.773459\\u003c/li\\u003e\\n \\u003cli\\u003eDalmazane M, Gallou-Guyot M, Compagnat M, Magy L, Montcuquet A, Billot M, et al. Effects on gait and balance of home-based active video game interventions in persons with multiple sclerosis: A systematic review. Mult. Scler. Relat. Disord. Elsevier B.V.; 2021. https://doi.org/10.1016/j.msard.2021.102928\\u003c/li\\u003e\\n \\u003cli\\u003eBorson S, Scanlan JM, Chen P, Ganguli M. The Mini-Cog as a screen for dementia: Validation in a population-based sample. J Am Geriatr Soc. 2003;51:1451\\u0026ndash;4. https://doi.org/10.1046/j.1532-5415.2003.51465.x\\u003c/li\\u003e\\n \\u003cli\\u003eGutierrez-Farewik EM, Bartonek \\u0026Aring;, Saraste H. Comparison and evaluation of two common methods to measure center of mass displacement in three dimensions during gait. Hum Mov Sci. 2006;25:238\\u0026ndash;56. https://doi.org/10.1016/j.humov.2005.11.001\\u003c/li\\u003e\\n \\u003cli\\u003eRusso Y, Berchicci M, Di Russo F, Vannozzi G. How do different movement references influence ERP related to gait initiation? A comparative methods\\u0026rsquo; assessment. J Neurosci Methods. Elsevier B.V.; 2019;311:95\\u0026ndash;101. https://doi.org/10.1016/j.jneumeth.2018.10.006\\u003c/li\\u003e\\n \\u003cli\\u003eRusso Y, Vannozzi G. Anticipatory postural adjustments in forward and backward single stepping: Task variability and effects of footwear. J Biomech. Elsevier Ltd; 2021;122. https://doi.org/10.1016/j.jbiomech.2021.110442\\u003c/li\\u003e\\n \\u003cli\\u003eQuijoux F, Nicola\\u0026iuml; A, Chairi I, Bargiotas I, Ricard D, Yelnik A, et al. A review of center of pressure (COP) variables to quantify standing balance in elderly people: Algorithms and open-access code*. Physiol. Rep. American Physiological Society; 2021. https://doi.org/10.14814/phy2.15067\\u003c/li\\u003e\\n \\u003cli\\u003eKoo TK, Li MY. A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for Reliability Research. J Chiropr Med. Elsevier USA; 2016;15:155\\u0026ndash;63. https://doi.org/10.1016/j.jcm.2016.02.012\\u003c/li\\u003e\\n \\u003cli\\u003eBland JM, Altman DG. Measurement STATISTICAL METHODS FOR ASSESSING AGREEMENT BETWEEN TWO METHODS OF CLINICAL MEASUREMENT. The Lancet. 1986;307\\u0026ndash;10.\\u003c/li\\u003e\\n \\u003cli\\u003evan Doorn PF, Geerse DJ, van Bergem JS, Hoogendoorn EM, Nyman E, Roerdink M. Gait Parameters Can Be Derived Reliably and Validly from Augmented Reality Glasses in People with Parkinson\\u0026rsquo;s Disease Performing 10-m Walk Tests at Comfortable and Fast Speeds. Sensors. Multidisciplinary Digital Publishing Institute (MDPI); 2025;25. https://doi.org/10.3390/s25041230\\u003c/li\\u003e\\n \\u003cli\\u003eKoop MM, Rosenfeldt AB, Owen K, Penko AL, Streicher MC, Albright A, et al. The Microsoft HoloLens 2 Provides Accurate Measures of Gait, Turning, and Functional Mobility in Healthy Adults. Sensors. MDPI; 2022;22. https://doi.org/10.3390/s22052009\\u003c/li\\u003e\\n \\u003cli\\u003eMancini M, Horak FB. Potential of APDM mobility lab for the monitoring of the progression of Parkinson\\u0026rsquo;s disease. Expert Rev. Med. Devices. Taylor and Francis Ltd; 2016. p. 455\\u0026ndash;62. https://doi.org/10.1586/17434440.2016.1153421\\u003c/li\\u003e\\n \\u003cli\\u003eZampieri C, Salarian A, Carlson-Kuhta P, Aminian K, Nutt JG, Horak FB. The instrumented timed up and go test: Potential outcome measure for disease modifying therapies in Parkinson\\u0026rsquo;s disease. J Neurol Neurosurg Psychiatry. BMJ Publishing Group; 2010;81:171\\u0026ndash;6. https://doi.org/10.1136/jnnp.2009.173740\\u003c/li\\u003e\\n \\u003cli\\u003eMic\\u0026oacute;-Amigo ME, Bonci T, Paraschiv-Ionescu A, Ullrich M, Kirk C, Soltani A, et al. Assessing real-world gait with digital technology? Validation, insights and recommendations from the Mobilise-D consortium. J Neuroeng Rehabil. BioMed Central Ltd; 2023;20. https://doi.org/10.1186/s12984-023-01198-5\\u003c/li\\u003e\\n \\u003cli\\u003eWinter DA. Human balance and posture control during standing and walking. Gait Posture. 1995;3:193\\u0026ndash;214.\\u003c/li\\u003e\\n \\u003cli\\u003eGatev P, Thomas S, Kepple T, Hallett M. Feedforward ankle strategy of balance during quiet stance in adults. Journal of Physiology. 1999.\\u003c/li\\u003e\\n \\u003cli\\u003eAppiah-Kubi KO, Galgon A, Tierney R, Lauer R, Wright WG. Concurrent vestibular activation and postural training recalibrate somatosensory, vestibular and gaze stabilization processes. PLoS One. Public Library of Science; 2024;19. https://doi.org/10.1371/journal.pone.0292200\\u003c/li\\u003e\\n \\u003cli\\u003eGarg H, Dibble LE, Schubert MC, Sibthorp J, Foreman KB, Gappmaier E. Gaze Stability, Dynamic Balance and Participation Deficits in People with Multiple Sclerosis at Fall-Risk. Anatomical Record. Blackwell Publishing Inc.; 2018;301:1852\\u0026ndash;60. https://doi.org/10.1002/ar.23852\\u003c/li\\u003e\\n \\u003cli\\u003eKeune PM, Young WR, Paraskevopoulos IT, Hansen S, Muenssinger J, Oschmann P, et al. Measuring standing balance in multiple sclerosis: Further progress towards an automatic and reliable method in clinical practice. J Neurol Sci. Elsevier B.V.; 2017;379:157\\u0026ndash;62. https://doi.org/10.1016/j.jns.2017.06.007\\u003c/li\\u003e\\n \\u003cli\\u003eKleiner AFR, Pacifici I, Vagnini A, Camerota F, Celletti C, Stocchi F, et al. Timed Up and Go evaluation with wearable devices: Validation in Parkinson\\u0026rsquo;s disease. J Bodyw Mov Ther. Churchill Livingstone; 2018;22:390\\u0026ndash;5. https://doi.org/10.1016/j.jbmt.2017.07.006\\u003c/li\\u003e\\n \\u003cli\\u003eSalarian A, Horak FB, Zampieri C, Carlson-Kuhta P, Nutt JG, Aminian K. ITUG, a sensitive and reliable measure of mobility. IEEE Transactions on Neural Systems and Rehabilitation Engineering. 2010;18:303\\u0026ndash;10. https://doi.org/10.1109/TNSRE.2010.2047606\\u003c/li\\u003e\\n \\u003cli\\u003eMellone S, Tacconi C, Chiari L. Validity of a Smartphone-based instrumented Timed Up and Go. Gait Posture. 2012;36:163\\u0026ndash;5. https://doi.org/10.1016/j.gaitpost.2012.02.006\\u003c/li\\u003e\\n \\u003cli\\u003eSzaflik P, Kaczmarczyk A, Zadoń H, Szefler-Derela J, Wasiuk-Zowada D, Nowakowska-Lipiec K, et al. Instrumented Timed Up and Go Test as a Tool to Early Detection of Gait and Functional Mobility Impairments in Multiple Sclerosis. J Clin Med. MDPI AG; 2026;15:679. https://doi.org/10.3390/jcm15020679\\u003c/li\\u003e\\n \\u003cli\\u003eZampieri C, Salarian A, Carlson-Kuhta P, Aminian K, Nutt JG, Horak FB. The instrumented timed up and go test: Potential outcome measure for disease modifying therapies in Parkinson\\u0026rsquo;s disease. J Neurol Neurosurg Psychiatry. BMJ Publishing Group; 2010;81:171\\u0026ndash;6. https://doi.org/10.1136/jnnp.2009.173740\\u003c/li\\u003e\\n \\u003cli\\u003eCaronni A, Picardi M, Scarano S, Malloggi C, Tropea P, Gilardone G, et al. Pay attention: you can fall! The Mini-BESTest scale and the turning duration of the TUG test provide valid balance measures in neurological patients: a prospective study with falls as the balance criterion. Front Neurol. Frontiers Media SA; 2023;14. https://doi.org/10.3389/fneur.2023.1228302\\u003c/li\\u003e\\n \\u003cli\\u003eFranchignoni F, Horak F, Godi M, Nardone A, Giordano A. Using psychometric techniques to improve the balance evaluation systems test: The mini-bestest. J Rehabil Med. 2010;42:323\\u0026ndash;31. https://doi.org/10.2340/16501977-0537\\u003c/li\\u003e\\n \\u003cli\\u003eHillel I, Gazit E, Nieuwboer A, Avanzino L, Rochester L, Cereatti A, et al. Is every-day walking in older adults more analogous to dual-task walking or to usual walking? Elucidating the gaps between gait performance in the lab and during 24/7 monitoring. European Review of Aging and Physical Activity. Springer Verlag; 2019;16. https://doi.org/10.1186/s11556-019-0214-5\\u003c/li\\u003e\\n \\u003cli\\u003eShah V V., Curtze C, Sowalsky K, Arpan I, Mancini M, Carlson-Kuhta P, et al. Inertial Sensor Algorithm to Estimate Walk Distance. Sensors. MDPI; 2022;22. https://doi.org/10.3390/s22031077\\u003c/li\\u003e\\n \\u003cli\\u003eTheodorou C, Velisavljevic V, Dyo V, Nonyelu F. Visual SLAM algorithms and their application for AR, mapping, localization and wayfinding. Array. Elsevier B.V.; 2022;15. https://doi.org/10.1016/j.array.2022.100222\\u003cstrong\\u003e\\u003c/strong\\u003e\\u003c/li\\u003e\\n\\u003c/ol\\u003e\"}],\"fulltextSource\":\"\",\"fullText\":\"\",\"funders\":[],\"hasAdminPriorityOnWorkflow\":false,\"hasManuscriptDocX\":true,\"hasOptedInToPreprint\":true,\"hasPassedJournalQc\":\"\",\"hasAnyPriority\":true,\"hideJournal\":false,\"highlight\":\"\",\"institution\":\"\",\"isAcceptedByJournal\":false,\"isAuthorSuppliedPdf\":false,\"isDeskRejected\":\"\",\"isHiddenFromSearch\":false,\"isInQc\":false,\"isInWorkflow\":false,\"isPdf\":false,\"isPdfUpToDate\":true,\"isWithdrawnOrRetracted\":false,\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"journal-of-neuroengineering-and-rehabilitation\",\"isNatureJournal\":false,\"hasQc\":true,\"allowDirectSubmit\":false,\"externalIdentity\":\"jner\",\"sideBox\":\"Learn more about [Journal of NeuroEngineering and Rehabilitation](http://jneuroengrehab.biomedcentral.com/)\",\"snPcode\":\"12984\",\"submissionUrl\":\"https://submission.nature.com/new-submission/12984/3\",\"title\":\"Journal of NeuroEngineering and Rehabilitation\",\"twitterHandle\":\"@BioMedCentral\",\"acdcEnabled\":true,\"dfaEnabled\":true,\"editorialSystem\":\"em\",\"reportingPortfolio\":\"BMC/SO AJ\",\"inReviewEnabled\":true,\"inReviewRevisionsEnabled\":true},\"keywords\":\"multiple sclerosis, augmented reality, digital biomarkers, gait analysis, wearable sensors, home assessment, Romberg, TUG, 6-minute walk test\",\"lastPublishedDoi\":\"10.21203/rs.3.rs-9424668/v1\",\"lastPublishedDoiUrl\":\"https://doi.org/10.21203/rs.3.rs-9424668/v1\",\"license\":{\"name\":\"CC BY 4.0\",\"url\":\"https://creativecommons.org/licenses/by/4.0/\"},\"manuscriptAbstract\":\"\\u003cp\\u003e\\u003cstrong\\u003eBackground:\\u003c/strong\\u003e Multiple sclerosis (MS) commonly affects balance and mobility, yet routine clinical tests may lack sensitivity to subtle change. Standardized mobility assessments such as the Timed Up and Go (TUG), six-minute walk test (6MWT), and Romberg balance test are widely used in neurorehabilitation but rely on manual timing or laboratory instrumentation. In this study we evaluated whether motor outcomes derived from augmented-reality (AR) glasses (Magic Leap 2) can provide valid, minimally instrumented measures of balance and gait.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eMethods:\\u003c/strong\\u003e Forty-six adults (26 people with MS; 20 healthy controls, HC) completed Romberg balance testing, TUG, and 6MWT while wearing a head-mounted AR device. AR outcomes were derived from AR glasses and processed using proprietary Strolll algorithms. Gold-standard reference measures were collected concurrently using 18-camera motion capture (100 Hz) and force plates for Romberg (1000 Hz). Primary outcomes were sway area (Romberg; 95% ellipse area), task duration and total distance (TUG/6MWT); secondary outcomes included mean gait velocity and step count (TUG/6MWT) as well as cadence, step length, and number of laps (6MWT). Concurrent validity was quantified using ICC(2,1) absolute agreement, Bland-Altman plots, mean absolute error and root mean square error.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eResults:\\u003c/strong\\u003e Agreement was excellent for primary outcomes, including TUG duration (ICC = 0.927–0.992) and 6MWT total distance (ICC = 0.966), with low absolute error relative to clinical variability. Secondary gait-derived metrics demonstrated good-to-excellent agreement (ICC range 0.750–0.980). AR-derived sway area showed poor-to-good agreement with force plate COP sway area across stance conditions. Agreement was preserved across healthy and clinical cohorts and across task variations.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eConclusions:\\u003c/strong\\u003e Head-mounted AR–derived outcomes demonstrated good-to-excellent concurrent validity against laboratory gold standards in MS and controls, with strongest performance for global mobility metrics (i.e., duration, distance, mean velocity) and more variable performance for step-derived measures (e.g. cadence, step length). These findings support AR-based assessment as a valid minimally instrumented approach, providing measurement performance consistent with gold-standard metrics for TUG and 6MWT (time, distance) and showing its potential for objectively monitoring disease progression and supporting rehabilitation in clinical and community settings.\\u003c/p\\u003e\",\"manuscriptTitle\":\"Motor and Balance Assessment in Multiple Sclerosis Using Augmented Reality: Concurrent Validation Against Laboratory Reference Measures\",\"msid\":\"\",\"msnumber\":\"\",\"nonDraftVersions\":[{\"code\":1,\"date\":\"2026-04-22 11:26:01\",\"doi\":\"10.21203/rs.3.rs-9424668/v1\",\"editorialEvents\":[{\"type\":\"communityComments\",\"content\":0},{\"type\":\"reviewerAgreed\",\"content\":\"144324439273949496404202490381934096288\",\"date\":\"2026-04-28T18:38:43+00:00\",\"index\":\"hide\",\"fulltext\":\"\"},{\"type\":\"reviewersInvited\",\"content\":\"\",\"date\":\"2026-04-28T10:34:32+00:00\",\"index\":\"\",\"fulltext\":\"\"},{\"type\":\"editorAssigned\",\"content\":\"\",\"date\":\"2026-04-20T14:13:42+00:00\",\"index\":\"\",\"fulltext\":\"\"},{\"type\":\"checksComplete\",\"content\":\"\",\"date\":\"2026-04-20T14:12:46+00:00\",\"index\":\"\",\"fulltext\":\"\"},{\"type\":\"submitted\",\"content\":\"Journal of NeuroEngineering and Rehabilitation\",\"date\":\"2026-04-15T08:50:00+00:00\",\"index\":\"\",\"fulltext\":\"\"}],\"status\":\"published\",\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"journal-of-neuroengineering-and-rehabilitation\",\"isNatureJournal\":false,\"hasQc\":true,\"allowDirectSubmit\":false,\"externalIdentity\":\"jner\",\"sideBox\":\"Learn more about [Journal of NeuroEngineering and Rehabilitation](http://jneuroengrehab.biomedcentral.com/)\",\"snPcode\":\"12984\",\"submissionUrl\":\"https://submission.nature.com/new-submission/12984/3\",\"title\":\"Journal of NeuroEngineering and Rehabilitation\",\"twitterHandle\":\"@BioMedCentral\",\"acdcEnabled\":true,\"dfaEnabled\":true,\"editorialSystem\":\"em\",\"reportingPortfolio\":\"BMC/SO AJ\",\"inReviewEnabled\":true,\"inReviewRevisionsEnabled\":true}}],\"origin\":\"\",\"ownerIdentity\":\"4038ce14-3380-44dc-b2c2-e71bb26606f5\",\"owner\":[],\"postedDate\":\"April 22nd, 2026\",\"published\":true,\"recentEditorialEvents\":[],\"rejectedJournal\":[],\"revision\":\"\",\"amendment\":\"\",\"status\":\"under-review\",\"subjectAreas\":[],\"tags\":[],\"updatedAt\":\"2026-04-28T10:39:00+00:00\",\"versionOfRecord\":[],\"versionCreatedAt\":\"2026-04-22 11:26:01\",\"video\":\"\",\"vorDoi\":\"\",\"vorDoiUrl\":\"\",\"workflowStages\":[]},\"version\":\"v1\",\"identity\":\"rs-9424668\",\"journalConfig\":\"researchsquare\"},\"__N_SSP\":true},\"page\":\"/article/[identity]/[[...version]]\",\"query\":{\"redirect\":\"/article/rs-9424668\",\"identity\":\"rs-9424668\",\"version\":[\"v1\"]},\"buildId\":\"XKTyCvWXoU3ODBz1xrDgd\",\"isFallback\":false,\"isExperimentalCompile\":false,\"dynamicIds\":[84888],\"gssp\":true,\"scriptLoader\":[]}","source_license":"CC-BY-4.0","license_restricted":false}