Effectiveness of action observation therapy based on virtual reality technology in the motor rehabilitation of paretic stroke patients: a randomized clinical trial

Stroke is a leading cause of adult disability, with more than 60% of survivors exhibiting upper limb motor deficits. Action observation therapy (AOT), grounded in the mirror neuron system (MNS), has shown promise for improving motor function when patients observe goal-directed actions and immediately imitate them. Virtual reality (VR) offers engaging, adaptable, and potentially home-based rehabilitation environments. This study hypothesizes that combining AO with VR (AO+VR) will be more effective than control observation of non-action scenes followed by VR (CO+VR) for improving upper limb function in chronic hemiplegic stroke patients. Secondary hypotheses include that baseline motor performance, cognitive level, and structural brain damage will correlate with the degree of improvement, and that AO+VR will induce greater functional plastic changes in MNS activity than CO+VR.

Prior clinical trials and meta-analyses have demonstrated the efficacy of AOT in improving upper limb function in stroke patients and other neurological conditions, including Parkinson's disease and unilateral cerebral palsy. The mirror neuron system (MNS) in humans involves regions such as the inferior parietal lobule (IPL), ventral premotor cortex (PMv), and inferior frontal gyrus (IFG), which are activated during both action execution and observation. Early trials (e.g., Ertelt et al.) reported functional improvement and increased cortical MNS activation maintained for weeks post-treatment. Franceschini et al. conducted randomized controlled trials showing AOT as an effective add-on to standard rehabilitation in acute stroke. VR-based rehabilitation has shown positive effects on upper limb motor function, with large studies (e.g., N=376) and Cochrane reviews supporting improvements over standard care. However, rigorous RCTs specifically testing combined AO followed by imitation within VR (AO+VR) are lacking, particularly in chronic stroke.

Design: Multicenter, allocation-concealed (waitlist-controlled), evaluator-blinded randomized controlled trial with two arms: AO+VR (experimental) vs CO+VR (control). Centers: Cardinal Ferrari Center (S. Stefano Riabilitazione, Parma, Italy) and Quarenghi Clinical Institute (Bergamo, Italy). Participants: Adults 3–18 months post-stroke with primarily motor symptoms, unilateral upper limb paresis, and residual movement ability. Inclusion: unilateral upper limb paresis; residual movement ability (MRC >2 and <4); active use from minimal to discrete; sufficient cooperation/cognitive understanding. Exclusion: severe cognitive impairment (MMSE <20); severe unilateral spatial neglect (Bells Test cutoff ≥50%); severe ideomotor apraxia; severe anosognosia; severe language comprehension deficits; severe untreated psychiatric disorders; sensory impairments hindering participation/uncorrected central visual deficits; drug-resistant epilepsy. MRI subgroup additional exclusions: insufficient cooperation for ~45 min neuroimaging; standard 3 T MRI contraindications. Setting: Quiet room; patient seated with arms on table at waist height; 32-inch monitor at 1.5 m; assisted by trained OT/PT. System: Khymeia Virtual Reality Rehabilitation System (VRRS) with 3D motion tracking (Polhemus Liberty), high-resolution monitor, and kinematic sensors for 3D tracking. Interventions: Experimental AO+VR—18 sessions (~1 h each; tolerance ±2; min 16, max 20), ~4 days/week over 5 weeks. Each session: observe 1.5-min videos of unimanual/bimanual actions (lateral perspective) then imitate actions at least 3 times within 3 min in VR using same objects, wearing 3D tracking sensors. Total of 52 exercises escalating in complexity (no-gravity sliding movements, simple/complex anti-gravity actions, functional bimanual tasks, complex daily life actions, obstacle avoidance/path following). Facilitation level (sensor sensitivity 1–10) adjusted per impairment; personalization based on baseline hand ability (minimal vs discrete). Each AO+VR session duration for exercises: 30–45 min plus setup changes. Control CO+VR—identical number and scheduling; observe 1.5-min naturalistic non-action videos; then perform same VR motor exercises prompted by therapist without prior action observation; thus purely motor execution without imitation component. Modifications: Temporary suspension if unable to perform VR; adverse events recorded; principal investigator decides termination. Adherence: Face-to-face reminder sessions at start and mid-treatment. Concomitant care: Usual medications and rehabilitation continue. Monitoring: Clinical Trial Quality Team (CTQT), University Hospital of Parma. Outcomes: Primary—Box and Block Test (BBT) administered by blinded assessor at baseline (T0), post-treatment within 72 h (T1), and 6-month follow-up (T2). Secondary—Modified Ashworth Scale (MAS), Motricity Index, Modified Rankin Scale (RS), Modified Barthel Index (mBI) at T0, T1, and T2. fMRI substudy: ~20 participants (10 per group) assessed pre (T0) and post (T1 within 72 h). Tasks: Action observation (LCD goggles; 72 4-s clips per condition; block design; object color balanced; rest fixation); Motor task (tool manipulation with paretic or both hands; control reaching; block design; rest fixation). MRI acquisition: 3 T GE MR750 Discovery, 32-channel head-coil; functional EPI (TR 2000 ms, TE 30 ms, 40 axial slices, slice thickness 3 mm + 0.5 mm gap, 64×64×37 matrix, FOV 205×205 mm², flip angle 90°, in-plane 3.2×3.2 mm²); T1 BRAVO (1×1×1 mm; TR 9700 ms; TE 4 ms; FOV 252×252 mm; 196 slices; flip angle 9°); 3D FLAIR CUBE (1×1×1 mm; TR 7000 ms; TE 118 ms; ETL 200 ms; FOV 256×256 mm; matrix 256×256; accel 2/2); DTI (b=1000 s/mm²; 64 directions; 8 b0; 45 axial slices; 3 mm; FOV 240×296 mm; matrix 80×80; TR 11000 ms; TE 92 ms). Randomization: Computer-generated, 1:1, stratified by hand use level (minimal vs discrete), permuted blocks of undisclosed sizes, sequence generated by independent researcher. Sample size: For BBT—clinically significant post-test change assumed 7 points; common SD 11.3; 10% dropout; 47 per group (N=94) to achieve 80% power. fMRI power: NeuroPower-tool indicates ≥9 participants per group for BOLD change with p<0.05 and 80% power (Bonferroni correction). Blinding: Participants/caregivers blinded to allocation; therapists not blinded; outcome assessors blinded. Data management: Encrypted, password-protected CRFs; DICOM transfer to PACS; alphanumeric ID codes; limited linkage access. Statistical analysis: SPSS v21; descriptive statistics; primary endpoint evaluated with independent samples t-test or Mann–Whitney if non-normal; intention-to-treat; rmANOVA mixed 2×3 (group between; time within: T0, T1, T2); missing data handled via linear mixed models (LMM) comparing with rmANOVA; secondary endpoints via LMM; alpha two-tailed p<0.05. fMRI analysis: SPM12 preprocessing (realignment, slice-timing, coregistration, normalization to MNI, smoothing); GLM; whole-brain t-contrast maps; ROI analyses in MNS areas; Student’s t-tests for BOLD differences; multiple comparison correction by FWE, significance p<0.001; additional PPI and DCM for connectivity changes. Ethics: Approved by AVEN (11 Nov 2020) and Bergamo (12 May 2021); informed consent required. Study status: Recruitment not started at submission.

This is a study protocol; no clinical results are reported. Planned parameters include: primary outcome Box and Block Test (BBT) with a clinically significant change defined as 7 points; assumed post-test SD 11.3; sample size N=94 (47 per group) for 80% power with 10% dropout. fMRI substudy power indicates ≥9 participants per group to detect BOLD changes with p<0.05 and 80% power (Bonferroni-corrected).

The study proposes that combining action observation with VR-based rehabilitation can enhance upper limb motor facilitation and functional recovery in chronic stroke, addressing limitations of traditional therapy. The AO+VR protocol is personalized to impairment level, incorporates ecologically valid bimanual daily actions, and leverages VR’s engaging, adaptable, and potentially home-based features. The inclusion of fMRI aims to elucidate MNS plasticity underlying motor improvements and to identify neural predictors of treatment response, including lateralization and activation extent in ipsilesional versus contralesional hemispheres. Findings could inform which patient profiles benefit most from AO+VR and support integration into standard rehabilitation and tele-rehabilitation.

This protocol outlines a rigorous multicenter RCT to evaluate AO+VR versus CO+VR for upper limb recovery in chronic stroke, with comprehensive clinical outcomes and an fMRI substudy to probe MNS-related plasticity. The anticipated contribution is evidence on the efficacy and neural mechanisms of AO+VR, informing personalized rehabilitation strategies and potential tele-rehabilitation applications. Future research directions include extending AO+VR to other neurological motor deficits and refining patient selection based on neuroimaging biomarkers.

Lesion localization affecting cortical areas associated with the mirror neuron system may limit the efficacy of AO-based interventions. Severity of motor impairment may constrain the feasibility of intensive training, necessitating facilitation and personalization. As a protocol, generalizability and actual treatment effects remain to be established pending trial completion.