Two adults with chronic stroke-related upper-limb impairment completed a long-term course of myoelectric virtual-reality training, known as MVR, with no clinical adverse events recorded. But the experiment involved only two people and was designed mainly to assess feasibility and safety; any functional gains were preliminary.
The system was tested in a very small group
Each session combined EMG recording, offline myoelectric pattern-recognition training and a posture-matching test aimed at wrist and hand movements on the affected side. The study was prospective, open-label and nonrandomized. Participant A followed a before-and-after design, while Participant B followed an ABA phase structure, with the MVR phase between two other phases. The phase order was not randomized, and assessment was not blinded or concealed.
Researchers screened 130 hospital records, identified five eligible contacts and found that three could not perform the EMG task. The two enrolled participants were a 52-year-old man, four years and one month after a right thalamic hemorrhage, and an 87-year-old woman, one year and six months after a left lacunar infarction.
Both participants completed the planned MVR-plus-rehabilitation intervention without discontinuation. Participant A completed 40 sessions and 1,368 posture-matching repetitions. Participant B completed 672 repetitions after the rehabilitation-alone phase was reduced to seven sessions. No clinical adverse events were recorded.
Movement measures sent mixed signals
The system's single-task accuracy, a measure of how often it identified the intended posture, rose from 86.8 to 93.5 for Participant A and from 83.3 to 91.8 for Participant B. Accuracy varied substantially, and most temporal changes were not statistically significant. Only Participant A's wrist palmar-flexion score showed the reported positive correlation, with a coefficient of 0.39 and a reported significance value of 0.018.
For the offline accuracy analysis, the score was based on the classifications needed to obtain 20 correct classifications, with 75% used as the reference score. Pearson correlations across MVR-plus-rehabilitation sessions were assessed using a 0.05 significance level.
On the Fugl-Meyer upper-extremity motor score, Participant A improved by 7 points, exceeding the estimated minimum clinically important difference. Participant B improved by 5 points, which did not exceed it. These were individual within-person changes in a design without a randomized comparator.
Gross manual dexterity also increased. Participant A's box-and-block score rose from 2 to 5 blocks between months 0 and 8. Participant B's score rose from 4 at month 0 to 12 at month 4, a 7-block increase from month 1 to month 4, before decreasing by 2 blocks at month 5.
Muscle-tone scores on the Modified Ashworth Scale improved by 1 point for Participant A and by 2 grades for Participant B during MVR plus rehabilitation, with Participant B's improvement maintained during carryover. Range-of-motion results differed: Participant A gained 15 degrees in wrist dorsiflexion and 10 degrees in finger flexion and extension, while Participant B had no significant overall change except for a 15-degree reduction in wrist palmar flexion during months 3 to 5.
Participants' reports offered a more mixed picture. Participant A described early dissociation between avatar and hand movements, then later reported less effort and fatigue. Participant B could perform posture matching but gave no clear subjective report because of dysarthria.
A feasibility signal, not an effectiveness result
The results do not show that MVR caused the observed changes. With two participants, no randomized external control and no blinded assessment, the study can show what changed within these cases, but not whether the training itself was responsible. The authors interpret long-term MVR plus conventional rehabilitation as feasible and generally safe, while stating that adequately powered controlled trials are needed before effectiveness conclusions can be drawn.
Paper data and sources
Original title: Long-term electromyography-based virtual reality training for chronic upper-limb motor paralysis after stroke: an exploratory single-case experimental design
Authors: Nagaoka Y, Suzuki S, Nakajima T et al.
Journal/Repository: Frontiers in rehabilitation sciences
Status: Peer-reviewed
First online: 2026-10-08
DOI: Not available
Original paper · Full text