PaperlayerView exact revision on arXiv

Source-linked AI summary

Towards the Threshold: A Fall-Risk Anchored Pareto Framework for Virtual Reality Gait Feedback Selection for Individuals with Multiple Sclerosis

Nafisa Anjum, John Quarles, M. Rasel Mahmud

arXiv:2609.37952v1cs.HC

TL;DR

VR gait studies in MS have often emphasized performance without accounting for the effort required, despite walking speed’s association with fall risk. This study compares eight feedback conditions using an external velocity target and burden measures. Spatial Auditory feedback had a favorable performance-burden trade-off, while the greatest gap closure came with the highest burden.

  • Problem

    VR rehabilitation studies have often emphasized performance improvements without systematically considering the cognitive and physical effort required, despite the association between reduced walking speed and fall risk in MS.

    Source: §1 · p. 1

  • Method

    The study evaluated eight VR feedback conditions in 34 adults with MS by measuring velocity-gap closure to an independently derived non-faller target alongside cognitive and physical burden, then applying Pareto analysis.

    Source: §1 · p. 1

  • Results

    Spatial Auditory closed 72.7% of the velocity gap with below-average burden; Multimodal achieved the greatest gap closure, 95.6%, but imposed the highest burden.

    Source: §6 · p. 10

  • Takeaways & Limitations

    The framework provides a practical method for evaluating VR rehabilitation conditions and guiding clinically meaningful protocol design for people with MS.

    Source: §6 · p. 10

    The 1.29 m/s target came from 41 MS participants with a small, nonsignificant faller–non-faller difference, so larger datasets with confirmed fall outcomes are needed to validate it as an individual rehabilitation target.

    Source: §5.6 · p. 9

Abstract

from arXiv · show

Reduced walking speed in people with multiple sclerosis (MS) is associated with an increased risk of falls. However, virtual reality (VR)-based rehabilitation studies often emphasize performance improvements without considering the cognitive and physical effort required to achieve them. This study introduces a threshold-anchored efficiency framework that jointly evaluates gait performance and overall effort. A normative walking-velocity target of \(T=1.29\) m/s was derived from the mean walking speed of non-fallers in an independent MS gait dataset and used as a clinically motivated reference point. Thirty-four adults with MS were evaluated across eight VR feedback conditions spanning unimodal, bimodal, and multimodal feedback. For each condition, we quantified the proportion of the velocity gap to the target that was closed and the associated cognitive and physical burden. Pareto efficiency analysis identified five non-dominated conditions: Static Visual, Spatial Auditory, Auditory+Visual, Auditory+Vibrotactile, and Multimodal, whereas Spatial Vibrotactile and Vibrotactile+Visual were dominated. Spatial Auditory showed a favorable performance-effort trade-off, closing 72.7% of the velocity gap while maintaining below-average burden. Multimodal feedback achieved the greatest gap closure (95.6%) but also imposed the highest burden. These findings demonstrate that greater gait improvement does not necessarily correspond to greater rehabilitation efficiency and provide a quantitative framework for comparing VR feedback strategies according to both performance gains and participant burden.

1 INTRODUCTION

Slower walking in MS is associated with greater fall risk, while VR feedback research has often prioritized performance without accounting for user burden. This study evaluates feedback strategies against an external velocity target and the effort required to approach it.

1 INTRODUCTION

  • VR feedback studies have often compared performance without systematically accounting for the cognitive and physical burden of achieving it.The study also notes that prior evaluations generally used within-study comparisons rather than clinically meaningful external benchmarks.Source: §1 · p. 1
  • The framework compares velocity-gap closure to an externally derived non-faller target with mental load and fatigue in 34 adults with MS.Pareto analysis identifies conditions that cannot be outperformed on both performance and burden.Source: §1 · p. 1
  • The framework applies external clinical benchmarking rather than comparing feedback conditions only against within-study means.The target was established from an independent MS gait dataset.Source: §1 · p. 1
  • Spatial Auditory feedback offered the most favorable performance-to-burden ratio among the evaluated options, while Spatial Vibrotactile and Vibrotactile+Visual were dominated.The study evaluated unimodal, bimodal, and multimodal conditions.Source: §1 · pp. 1–2
  • Pareto analysis is proposed as a condition-selection tool for identifying feedback that advances patients toward a safe velocity target without disproportionate burden.The resulting frontier is intended to support selection among dominant conditions.Source: §1 · p. 2

2 BACKGROUND AND RELATED WORK

The background connects MS gait impairment and fall risk with VR’s potential to guide walking, while noting that multisensory feedback can impose burden. It motivates evaluating rehabilitation options across performance and effort together.

2.1 Gait Impairment and Fall Risk in Multiple Sclerosis

  • In MS, gait impairment commonly includes reduced walking speed, and speed below normative benchmarks is associated with increased fall frequency and other functional consequences.Walking speed is described as a clinically sensitive gait indicator and is the primary outcome of the Timed 25-Foot Walk test.Source: §2.1 · p. 2

2.2 VR for Gait Rehabilitation

  • VR can support gait rehabilitation, but immersion can also disrupt gait through restricted vision, latency, and altered motion cues.The cited work reports both rehabilitation benefits and gait changes associated with visual-vestibular mismatch.Source: §2.2 · p. 2

2.3 Assistive Feedback in VR Walking

  • Auditory and vibrotactile cues have shown benefits in gait or balance, while the proportionality of multimodal performance gains to added burden remains unclear.Prior work reports larger gait improvements from auditory-vibrotactile combinations than either channel alone, alongside concern about greater cognitive and physical load.Source: §2.3 · p. 2

2.4 Efficiency Frameworks in Rehabilitation and HCI

  • Formal rehabilitation frameworks comparing multiple criteria remain rare, and Pareto analysis can identify efficient trade-offs as well as dominated conditions.The paper applies this approach to performance improvement and user burden as a basis for selecting rehabilitation strategies.Source: §2.4 · pp. 2–3

3 METHODOLOGY

The methodology evaluates feedback conditions against gait and burden hypotheses in adults with MS, using instrumented walking measures and a VR system that can deliver auditory, tactile, and visual cues.

3.1 Hypotheses

  • The study tests whether VR baseline walking is slower than non-VR walking, whether feedback advances velocity toward T, whether gain and burden vary proportionally, and whether a robust Pareto frontier exists.It also hypothesizes that the frontier remains similar when mental load and fatigue are analyzed separately.Source: §3.1 · p. 3

3.2 Participants

  • Thirty-four adults with MS and mobility impairment participated; eligibility required independent walking with or without an assistive device.Participants with other neurological or orthopedic conditions independently affecting gait were excluded.Source: §3.2 · p. 3

3.3.1 Apparatus and Virtual Environment

  • Walking velocity was recorded with a GAITRite instrumented walkway, while the VR environment used an HTC Vive Pro and synchronized gait data with the simulation.Participants were secured in a suspension harness during walking tasks.Source: §3.3.1 · p. 3

3.3.2 Multimodal Feedback System

  • The feedback system combined distributed vibrotactile actuators with spatialized audio cues and static visual anchors.The tactile hardware included a vest, forearm sleeves, and a forehead device.Source: §3.3.2 · p. 3

3.4 Experimental Feedback Conditions

Eight VR feedback conditions combined visual, auditory, and vibrotactile cues in different configurations, alongside VR and non-VR no-feedback baselines. The designs tested how sensory guidance could shape walking performance.

3.4.1 Auditory Feedback

  • Spatial audio delivered directional guidance through headphones, using white noise intended to avoid semantic or rhythmic dual-task interference.Audio was generated with the Google Resonance Audio SDK and localized in three dimensions.Source: §3.4.1 · p. 4

3.4.2 Vibrotactile Feedback

  • Vibrotactile cues mapped head orientation and body position to vest, sleeve, and forehead tactors.Tactile signals were synchronized with audio where applicable, and vibration intensity was adjusted for participant comfort.Source: §3.4.2 · p. 4

3.4.3 Visual Feedback

  • The visual condition used a head-fixed anchor of four corner markers and a central crosshair as a stable reference for stride alignment.The anchor moved with the participant’s head rather than staying fixed in the virtual world.Source: §3.4.3 · p. 4

3.4.4 Vibrotactile + Visual Feedback

  • Vibrotactile+Visual combined spatial tactile signals with the static visual anchor to examine walking regularity and gait deviations.The stated goal was comparison with either cue alone.Source: §3.4.4 · p. 4

3.4.5 Auditory + Visual Feedback

  • Auditory+Visual paired directional spatial audio with the stabilizing visual frame.Source: §3.4.5 · p. 4

3.4.6 Auditory + Vibrotactile Feedback

  • Auditory+Vibrotactile delivered synchronized directional cues through hearing and touch to evaluate multisensory reinforcement.The setup was intended to test whether combined cues provided stronger corrective guidance for walking stability than either modality alone.Source: §3.4.6 · p. 4

3.4.7 Multimodal Feedback

  • Multimodal feedback combined auditory, vibrotactile, and visual cues, with walking speed, stride consistency, and trajectory deviation analyzed.It was designed to test whether trimodal feedback offered advantages over unimodal or bimodal conditions.Source: §3.4.7 · p. 4

3.4.8 No Feedback (Non-VR and VR Baseline)

  • The VR baseline included the headset and equipment without feedback; the non-VR baseline represented walking without VR equipment.These conditions provided assisted-environment and natural-walking references, respectively.Source: §3.4.8 · p. 4

3.5 Subjective Burden Assessment

Subjective burden was assessed using mental-load and fatigue ratings, then summarized by averaging their grand z-scores. The composite showed good internal consistency.

3.5 Subjective Burden Assessment

  • Mental load and fatigue were each rated from 0 to 10, and only post-condition ratings entered the analysis.Mental load captured perceived cognitive demand; fatigue captured perceived physical tiredness.Source: §3.5 · p. 4
  • The burden composite averaged grand z-scores for mental load and fatigue across participants and conditions.For condition c, z(MLc) and z(Fc) represent standardized mental-load and fatigue ratings; the grand mean and standard deviation were computed across all participants and conditions.Source: §3.5 · pp. 4–5
  • Cronbach’s alpha was 0.86, indicating good internal consistency and supporting equal weighting of the two burden measures.Source: §3.5 · p. 5

3.6 Normative Velocity Threshold

The threshold was derived from laboratory walking data for adults with MS classified by prior-year fall history. Because velocity poorly discriminated fallers from non-fallers, the study used a normative benchmark rather than a diagnostic cutoff.

3.6.1 Reference Dataset

  • The reference dataset contained laboratory and home gait data from 41 community-dwelling adults with MS; only laboratory sessions were used to match the study’s controlled walking conditions.Source: §3.6.1 · p. 5

3.6.2 Fall Label Extraction and Participant Matching

  • Fall-history labels classified 22 participants as fallers and 19 as non-fallers, matched to gait measurements for all 41 participants.Source: §3.6.2 · p. 5

3.6.3 Group-Level Velocity Analysis

  • Non-fallers averaged 1.29 m/s (SD = 0.17, n = 19), versus 1.20 m/s (SD = 0.23, n = 22) for fallers; the 0.09 m/s difference was not statistically significant.The independent-samples t-test yielded p = .16 and the Mann-Whitney U test p = .19; Figure 4 presents the velocity distribution.Source: §3.6.3 · p. 5, Figure 4 · p. 5

3.6.4 Evaluation of a Diagnostic Threshold

  • Velocity classification had AUC 0.62, indicating weak discrimination, so a ROC-derived diagnostic cutoff was not adopted.Source: §3.6.4 · p. 5

3.6.5 Threshold Definition

  • The primary threshold T was the non-faller mean, 1.29 m/s; it represents typical non-faller walking speed, not a diagnostic or guaranteed fall-prevention cutoff.A lower bound one standard deviation below the mean was used only for sensitivity checks; movement toward T was interpreted as movement toward safer functional walking speed.Source: §3.6.5 · p. 5

3.7 Gap Closure Computation

Gap closure was assessed against the normative threshold and the VR no-feedback baseline. The non-VR baseline was excluded because it was not a VR feedback design option.

3.7 Gap Closure Computation

  • For each feedback condition, the velocity gap was computed relative to threshold T using that condition’s mean walking velocity across all 34 participants.Source: §3.7 · p. 5
  • Gap closure was defined relative to the VR Baseline; a positive value means the condition reduced the gap to T more than VR walking without feedback.The percentage of the gap closed was also calculated.Source: §3.7 · p. 5
  • The Non-VR Baseline was excluded from efficiency analysis because it was not a VR feedback design option.Source: §3.7 · p. 5

3.8 Pareto Efficiency Analysis

Pareto analysis identifies feedback options that balance velocity-gap closure against burden, while secondary measures characterize participants’ experience and perceived balance.

3.8 Pareto Efficiency Analysis

  • A condition is dominated when another condition performs at least as well with lower burden, and the non-dominated options form the Pareto frontier.The VR Baseline is excluded because its gap closure is zero by definition.Source: §3.8 · p. 6
  • The efficiency ratio summarizes gap closure per unit of burden and ranks frontier conditions; sensitivity analyses repeat the analysis using mental load or fatigue alone.A constant denominator offset keeps values positive across the observed burden range.Source: §3.8 · p. 6

3.9 Secondary Measures

  • Presence was assessed with a 0–10 single-item rating of immersion in the virtual walking environment.The measure followed the approach in.Source: §3.9 · p. 6
  • Usability was assessed with a Likert-scale response to whether participants considered the system usable.The item was intended to capture perceived practicality of feedback conditions.Source: §3.9 · p. 6
  • The study also recorded balance confidence with the ABC Scale and monitored cybersickness with the SSQ before and after participation.ABC scores were categorized as high, moderate, or low; the SSQ monitored nausea, oculomotor strain, and disorientation.Source: §3.9 · p. 6

3.10 Study Procedure

Participants completed matched timed-walking tasks in real and virtual environments, with safety, feedback settings, and condition order controlled.

3.10 Study Procedure

  • Feedback settings were calibrated individually for comfort, and the real-versus-virtual task comparison is shown in Figure 5.Figure 6 depicts harnessed GAITRite walking in both environments.Source: §3.10 · p. 6, Figure 5 · p. 6, Figure 6 · p. 6
  • Participants completed three comfortable-pace timed-walking trials on the GAITRite walkway in the real environment while secured by an overhead safety harness.They performed 180-degree turns at each end and stepped off the mat during turns because GAITRite does not record turning movements.Source: §3.10 · p. 6
  • The timed-walking task was repeated in VR across feedback conditions and a no-feedback baseline, with three trials per condition presented in counterbalanced order.Mental load, fatigue, presence, and usability ratings were collected around each condition.Source: §3.10 · p. 6

4 RESULTS

Across the eight feedback conditions, five formed the Pareto frontier and two were dominated; gap closure and burden varied, and the frontier remained stable under alternative burden measures.

4.1 Statistical Analysis

  • Walking velocity differed significantly across the nine conditions, and post-hoc paired tests examined condition pairs; ordinal subjective measures were analyzed with non-parametric tests.The repeated-measures ANOVA reported F(8,264) = 171.30, p < .001, η2p = 0.839.Source: §4.1 · pp. 6–7

4.2 VR-Induced Gait Disruption

  • VR Baseline walking velocity was 1.0797 m/s versus 1.2926 m/s in Non-VR Baseline, a significant reduction of 0.2129 m/s before feedback.Cadence, step length, and stride length decreased, while step time and swing time increased under VR Baseline.Source: §4.2 · p. 7

4.3 Normative Threshold Derivation

  • The target T = 1.29 m/s uses the reference dataset’s non-faller mean as a directional benchmark, not a diagnostic cut-point.The groups showed weak, non-significant separation (d = 0.44, AUC = 0.62, p = .16).Source: §4.3 · p. 7

4.4 Walking Velocity and Gap Closure

  • Gap closure ranged from 22.7% for Static Visual to 95.6% for Multimodal, and the Non-VR Baseline exceeded the target by 0.004 m/s.Conditions with similar raw velocities did not necessarily rank similarly in target-relative gap closure.Source: Table 4 · p. 7, §4.4 · p. 7, Figure 8 · p. 7

4.5 Subjective Burden Across Conditions

  • Burden increased with feedback complexity, reaching its highest composite value in Multimodal, which also produced significantly higher mental load and fatigue than unimodal conditions.The mental-load and fatigue composite had Cronbach’s alpha of 0.86; presence, usability, and cybersickness showed no detectable condition differences.Source: §4.5 · p. 7, Table 5 · p. 8

4.6 Pareto Efficiency Analysis

  • Five feedback conditions formed the Pareto frontier, while Spatial Vibrotactile and Vibrotactile+Visual were dominated; Spatial Auditory had the highest efficiency ratio and Multimodal the greatest gap closure but highest burden.The reported efficiency ratio for Spatial Auditory was 0.086; Multimodal closed 95.6% of the velocity gap.Source: Table 6 · p. 8, §4.6 · p. 8

4.7 Sensitivity Analysis

  • The Pareto frontier was unchanged when burden was measured by the composite, mental load alone, or fatigue alone.The five frontier and two dominated conditions were the same in all three analyses.Source: §4.7 · p. 8

5 DISCUSSION

The discussion frames VR immersion as a gait challenge and evaluates feedback by both progress toward the target and participant burden. Spatial Auditory offers the strongest efficiency trade-off, while Multimodal maximizes gap closure at the highest burden.

5.1 The Cost of Immersion

  • VR Baseline walking was approximately 0.21 m/s slower than Non-VR Baseline and 0.21 m/s below the normative threshold.The authors describe unassisted VR walking as lying in a velocity range associated with increased fall risk.Source: §5.1 · p. 8

5.2 Advancing Toward the Threshold: The Clinical Case for VR Feedback in MS

  • All eight feedback conditions reduced the velocity gap relative to VR Baseline, with closure ranging from 22.7% to 95.6%.Multimodal reached 1.279 m/s, within 0.01 m/s of the threshold; Static Visual closed the least gap but had the lowest burden among feedback conditions.Source: §5.2 · p. 8

5.3 Performance Versus Burden: Why Effectiveness Alone Is Not Enough

  • Performance gains and subjective burden did not rise together: Spatial Auditory closed 72.7% of the gap at below-average burden, while Vibrotactile+Visual closed 59.4% at higher burden.The authors suggest sensory-channel reliability may help explain why Spatial Auditory outperformed Vibrotactile+Visual, but present this as a possible explanation.Source: §5.3 · pp. 8–9

5.4 The Pareto Frontier as a Clinical Decision Tool

  • Spatial Vibrotactile and Vibrotactile+Visual were dominated in this implementation; Spatial Auditory had the highest efficiency ratio, while Multimodal delivered the most gap closure at the highest burden.Static Visual, Auditory+Visual, and Auditory+Vibrotactile provide intermediate frontier trade-offs.Source: §5.4 · p. 9, Figure 9 · p. 9

5.5 Robustness of the Efficiency Structure

  • The same five frontier and two dominated conditions emerged using composite burden, mental load alone, or fatigue alone.The authors report that Likert-scale ratings were sufficient to identify the same conditions.Source: Table 7 · p. 8, §5.5 · p. 9

5.6 Limitations and Future Directions

  • The study’s findings reflect one experimental session, a threshold derived from 41 MS participants without significant faller/non-faller separation, and two single-item burden measures.The authors call for larger datasets with confirmed fall outcomes and longitudinal studies to assess sustained real-world walking effects; whether the Pareto structure generalizes beyond MS remains open.Source: §5.6 · p. 9

6 CONCLUSION

The framework evaluates eight VR feedback conditions by velocity-gap closure toward a normative fall-risk threshold and their cognitive and physical costs. All conditions improved walking relative to the VR baseline, but performance gains and burden did not align directly.

6 CONCLUSION

  • All eight feedback conditions improved walking relative to the VR baseline, closing 22.7% to 95.6% of the velocity gap.Multimodal feedback achieved the greatest gap closure, 95.6%.Source: §6 · pp. 9–10
  • VR without feedback reduced walking velocity by approximately 0.21 m/s compared with real-world walking, placing participants further below the threshold.The normative fall-risk threshold was T = 1.29 m/s.Source: §6 · p. 9
  • Five conditions were Pareto-frontier options and two were dominated; Spatial Auditory closed 72.7% of the velocity gap with below-average burden.Spatial Auditory showed the highest efficiency among frontier conditions.Source: §6 · p. 10
  • Greater gait improvement did not necessarily correspond to greater rehabilitation efficiency because performance gain and user burden were not directly aligned.Multimodal feedback had the highest burden despite achieving the greatest gap closure.Source: §6 · p. 10
Loading 2609.37952v1…