Digital Outcome Assessment Using AI Active Gaming and Motion Capture in Friedreich Ataxia

Design and rationale STEP-OUT FA is an observational, prospective cohort study (STROBE-compliant) developing and validating a digital clinical outcome assessment (dCOA) for Friedreich ataxia. It addresses a well-documented limitation in FA therapeutic development: existing semi-quantitative clinical outcome assessments are affected by floor and ceiling effects, limited responsiveness across disease stages, and poor patient-perceived relevance, which have contributed to repeated failures of drug trials to meet primary endpoints. The study evaluates whether motor-performance metrics captured during gamified, semi-immersive virtual-reality (VR) tasks, combined with markerless AI motion analysis, can serve as reliable, valid, and responsive outcome measures.

Measurement platform The platform is delivered through a standard laptop and integrates three components: (1) FA-adaptable VR gamified tasks (e.g., ball catching, target reaching) with adjustable difficulty; (2) automated extraction of gaming-performance metrics; and (3) a single-camera, patented AI-driven markerless motion-capture system for 3D pose reconstruction (DigiTherapix Ltd), used to derive biomechanical/kinematic metrics. The gaming environment is provided via the MoveHero platform (University of São Paulo).

Laboratory validation (healthy controls and FA participants) Validation is conducted at the University of Exeter VSimulator facility against a gold-standard, marker-based full-body motion-capture reference. During a single laboratory visit (\~2 hours), participants complete a familiarisation period followed by a \~20-minute session of four movement-based gaming tasks. Concurrent signals are recorded throughout: full-body optical motion capture (reflective markers on shoulders, elbows, knees, ankles, forehead), continuous heart-rate variability via a Delsys ECG sensor (baseline, task, recovery), and functional near-infrared spectroscopy (fNIRS) over the forehead to index cortical activation. Perceived exertion and fatigue (Borg RPE and Borg VAS Fatigue) are collected after each task, and a short semi-structured interview captures usability and acceptability feedback (audio-recorded for later transcription). FA participants additionally complete clinical characterisation at baseline.

Home-based longitudinal phase (FA participants only) To evaluate test-retest reliability and responsiveness, FA participants enter a 12-week home phase with monthly self-administered or remotely supported gaming sessions (\~20-30 minutes) at Weeks 4, 8, and 12, each accompanied by a brief telehealth check-in with a physiotherapist to confirm continued consent, provide support, and monitor adverse events. Data are captured automatically and remotely. A remote follow-up assessment at Week 12 repeats clinical and patient-reported measures and includes a short interview.

Analytic approach Concurrent validity is assessed via correlations between digital metrics and established clinical scales; reliability via intraclass correlation coefficients across repeated sessions; and responsiveness via within-subject change and effect sizes, with longitudinal mixed-effects modelling of dCOA trajectories across the home-phase time points. Analyses characterise which individual or composite metrics (gaming-derived versus biomechanical) show optimal measurement properties and best discriminate disease severity. As an exploratory feasibility and validation study establishing proof of concept and psychometric properties, no formal power calculation was performed; the sample size was justified on precision for reliability estimation and on precedent from rare-disease dCOA validation studies.

Collaborators and data governance The AI motion-analysis platform is provided by DigiTherapix Ltd (UK) and the gaming platform by the University of São Paulo (Brazil); formal sub-contractor and data-sharing agreements govern intellectual property and data governance. No personally identifiable participant data are transferred outside the UK. Oxford University Hospitals acts as a Participant Identification Centre only.

• Laboratory validation session (VSimulator) (Device): Single laboratory visit at the University of Exeter VSimulator in which VR active gaming tasks and single-camera AI-driven markerless motion analysis are performed concurrently with gold-standard marker-based full-body motion capture, functional near-infrared spectroscopy (fNIRS), and continuous heart-rate variability (Delsys ECG sensor). Performed once.
• Home-based active gaming assessment (Device): Monthly home-based sessions (Weeks 4, 8, 12) in which participants perform the VR active gaming tasks with concurrent single-camera AI-driven markerless motion analysis via a loaned laptop, with automated remote data capture. No marker-based motion capture, fNIRS, or ECG.

Primary outcomes

• Concurrent validity of game-based movement performance against the mFARS - Friedreich ataxia cohort. [Baseline laboratory visit (Week 0)]: Spearman correlation coefficient (ρ; unitless) between game-based movement accuracy (Absolute Error), precision (Variable Error), reaction time, movement time and the modified Friedreich's Ataxia Rating Scale (mFARS) total score (range 0-93; higher = greater neurological impairment), in the Friedreich ataxia cohort.
• Criterion validity of AI markerless motion capture versus marker-based motion capture - Healthy Controls and Friedreich Ataxia [Baseline laboratory visit (Week 0)]: Agreement for upper and lower body angular velocity, joint angles and functional range of motion measured concurrently by the single-camera AI markerless system and the gold-standard marker-based motion-capture system.
• Comparison of the digital COA metrics between healthy control and Friedreich ataxia cohort. [Baseline laboratory visit (Week 0)]: Use of VR game-derived digital metrics (reaction time, movement time, movement accuracy and precision, and AI-derived kinematic metrics such as angular velocity, joint angles and Functional Range of Motion) to compare healthy control and Friedreich ataxia cohort.
• Responsiveness of digital COA metrics over time [Weeks 0, 4, 8, and 12]: Within-subject change scores and effect sizes for digital metrics, analysed with longitudinal mixed-effects modelling of dCOA trajectories across the home-based phase.
• Patient Global Impression of Change (PGIC) [Week 12]: Patient Global Impression of Change, 7-point ordinal scale (1 = very much improved to 7 = very much worse; higher = greater perceived worsening), in the FA cohort.
• Platform acceptability and usability [Weeks 0, 4, 8, and 12]: System Usability Scale (SUS) total score (range 0-100; higher = better usability; ≥70 acceptable, ≥80 excellent).

Secondary outcomes

• Cortical activation during gaming tasks (fNIRS) - healthy control and Friedreich ataxia cohort. [Baseline laboratory visit (Week 0)]: Changes in cortical haemodynamic response (oxygenated and deoxygenated haemoglobin) measured by functional near-infrared spectroscopy during the gaming session.
• Autonomic response during gaming tasks (heart-rate variability) - healthy control and Friedreich ataxia cohort. [Baseline laboratory visit (Week 0)]: Heart-rate variability indices derived from continuous ECG recording (Delsys sensor) at baseline, during, and after the gaming session.
• Perceived exertion (Borg CR10) - healthy control and Friedreich ataxia cohort. [Baseline laboratory visit (Week 0)]: Borg Rating of Perceived Exertion score (range 1-10; higher = greater exertion) after each gaming task.
• Fatigue (Borg VAS Fatigue) - healthy control and Friedreich ataxia cohort. [Baseline laboratory visit (Week 0)]: Visual Analogue Scale (VAS) for fatigue (range 0-10; higher = greater fatigue) after each gaming task.
• 9-Hole Peg Test completion time - healthy control and Friedreich ataxia cohort. [Baseline laboratory visit (Week 0)]: Time to complete the 9-Hole Peg Test (9HPT), measured in seconds (higher = worse manual dexterity; no fixed maximum), averaged across the dominant and non-dominant hands, in the Friedreich ataxia cohort.
• Gait quality: proportion of good steps (Heel2Toe sensor) [Baseline laboratory visit (Week 0)]: Proportion of "good steps" - steps in which the heel strikes first, detected by the Heel2Toe wearable sensor from ankle sagittal-plane angular velocity, expressed as a percentage of total steps taken during a standardised walking task (range 0-100%; higher = better gait quality), in the Friedreich ataxia cohort.

Sex: All, ages: 12 Years to 50 Years, accepts healthy volunteers

Friedreich ataxia group:

* Aged 12 to 50 years
* Genetically confirmed Friedreich ataxia
* Able to sit or stand independently for 5 minutes
* Able to provide informed consent (or assent with parental consent if under 16 years)
* Stable medication regimen (no changes in the past 4 weeks)

Healthy control group:

* Aged 12 to 50 years
* No history of genetically confirmed Friedreich ataxia
* Able to sit or stand independently for 5 minutes
* Able to provide informed consent (or assent with parental consent if under 16 years)

Exclusion Criteria:

* Friedreich ataxia group:
* Aged 11 years or younger, or 51 years or older
* Severe cognitive impairment precluding task comprehension
* Active seizure disorder, or vestibular dysfunction causing nausea with non-immersive VR
* Currently participating in an interventional clinical trial
* No reliable internet connection (where a digital inclusion pack cannot resolve this)
* Any reason precluding safe participation in moderate-intensity exercise.

Healthy control group:

* Aged 11 years or younger, or 51 years or older
* Genetically confirmed Friedreich ataxia
* Any reason precluding safe participation in moderate-intensity exercise.

• Helen Dawes, Principal investigator, University of Exeter
• Andrea Nemeth, Study director, University of Oxford