This study included 51 participants with upper limb functional deficits caused by stroke, who presented at Samsung Medical Center of Seoul and Pusan National University Yangsan Hospital of Yangsan-si, Republic of Korea. All participants were eligible for inclusion if they met the following requirements: (1) age between 20 and 85 years, (2) > 3 weeks and < 3 months after stroke onset, (3) active range of motion (ROM) in the wrist > 10 degrees, and (4) unilateral upper limb deficit with a Fugl-Myer Assessment score > 22. Exclusion criteria were (1) history of preexisting neurological or psychiatric disorder, (2) multiple or bilateral stroke lesions, (3) Korean Mini-Mental State Exam (K-MMSE) score < 17, (4) aphasia, and (5) pregnancy. Ethics approval was granted by the Pusan National University Yangsan Hospital Ethics Committee and written informed consent was obtained from all participants before the study. This study was retrospectively registered at ClinicalTrials.gov.
A randomized controlled trial was performed to test the effectiveness of hand motor training with the RAPAEL® Smart Glove digital system and game-based VR in subacute stroke patients. Eligible participants were randomly placed in either the experimental group (hand motor training with the RAPAEL® Smart Glove digital system) or control group (conventional OT for the same amount of time as the experimental group) by a research administrator using a random number table after baseline assessment. All participants were assigned a code number.
RAPAEL® Smart Glove digital system
The RAPAEL® Smart Glove digital system was designed to induce neuroplasticity for hand function and has two types of embedded sensors to collect information on individual motions in real-time. By applying a ‘Learning Schedule Algorithm’ to game-like exercises, the RAPAEL® Smart Glove can create ADL-related tasks compatible with an individual’s function level. The system provides information about a patient’s current condition, exercise progress, and functional improvement by analyzing the active ROM.
All participants were treated with 20 intervention sessions over 4 weeks: 5 times per week, 1 hour per day. The experimental group received game-based VR hand motor training with the RAPAEL® Smart Glove digital system for a total of 20 sessions at 5 sessions per week for 4 weeks. If participants missed any training during the intervention period, additional sessions were offered at another time during the week or during an optional additional week at the end of the intervention period. In each VR game, the participants were required to successfully perform tasks related to a specific intended movement to obtain a high score.
In the training protocol, the average time per session was 1 hour, divided into 30 min with the VR training program and 30 min of conventional OT. The intervention structure was customized to each participant’s hand function level. As the session progressed, the training intensity gradually increased by changing the VR game level. The control group had 1-hour sessions of conventional OT alone without VR hand motor training.
We performed the following assessments before intervention (T0), immediately after the intervention (T1), and 4 weeks after the intervention (T2).
Primary outcome: motor function
An occupational therapist performed upper extremity Fugl-Meyer assessment (UFMA) for motor impairment of the affected side and the Jebsen-Taylor hand function test (JTT) at T0, T1, and T2. Primary outcome was differences of these motor function scores between T1 and T0, T2 and T0. The UFMA consists of 33 items (3-point ordinal scale and range, 0–66), with higher scores indicating less impairment. The JTT assesses hand function according to ADL with a series of 7 timed subtests, including writing, simulated page turning, picking up small objects, simulated feeding, stacking checkers, picking up large light objects, and picking up large heavy objects. In original JTT, the subtest is considered missing if patient cannot complete the subtest within a certain amount of time.. In order to overcome this limitation of original JTT scoring system, we adopted the modified scoring system in this study as presented in the previous study. According to this modification, each subtest scored from 0 to 15 and the total score is the sum of each subtest scores ranged from 0 to 105) .
Secondary outcome: cortical activation changes in the motor cortical regions
To investigate cortical activation by changes of oxygenated hemoglobin (OxyHb), we used the NIRSscout® system (NIRx Medical Technology, Berlin, Germany), which is a multi-modal, compatible, functional near-infrared spectroscopy system (fNIRS) platform. This system has many optodes consisting of 16 sources and 16 detectors, which cover the sensorimotor cortex (SMC), premotor cortex (PMC), and supplementary motor area (SMA) using 45 channels of interest. The NIRSscout® uses two different wavelengths (760 nm and 850 nm) with a sampling rate of 3.91 Hz. The optodes were positioned according to the international 10/20 system, and the channel distance (i.e., distance between the source and detector) was 3.0 cm.
Changes in OxyHb concentration over the ipsilesional primary motor cortex is he secondary outcome which was analyzed by the NIRS-SPM (Near Infrared Spectroscopy-Statistical Parametric Mapping)  software package in MATLAB (The Mathworks, USA). To investigate cortical activity in the affected side of the brain, left brain lesions were flipped from left to right in the data preprocessing stage so all included lesions were set on the right. We used a modified Beer-Lambert law to calculate OxyHb level following change in cortical concentration . The international 10/20 system was used to position optodes with the cranial vertex (Cz) located beneath the 1st source. The nasion, left ear, right ear, and inion were identified in each subject. A stand-alone application was used for spatial registration of the 49 functional channels on the Montreal Neurological Institute brain.
Gaussian smoothing with a 2s full width at half maximum (FWHM) was applied to correct noise from the fNIRS system. A wavelet discrete cosine transform (DCT)-based detrending algorithm was used to correct signal distortion due to breathing or movement, and a general linear model (GLM) analysis with a canonical hemodynamic response curve was then performed to model the hypothesized OxyHb response under the experimental conditions . To investigate changes in cortical activation during wrist and hand movements, we selected 5 regions of interest (ROIs) defined by Brodmann area (BA) or anatomical markers: primary SMC (BA 1, 2, 3, and 4), PMC (BA 6), SMA (anterior boundary: vertical line to the anterior commissure, posterior boundary: anterior margin of primary SMC, medial boundary: midline between the right and left hemispheres, lateral boundary: 15 mm lateral to the midline between the right and left hemispheres).
All statistical analyses were performed with SPSS version 22.0 (IBM, Armonk, N.Y., USA), and the significance level was set at 0.05. The Shapiro-Wilk test was used to confirm that all outcome variables were normally distributed. The independent t-test for continuous parameters, Mann-Whitney U test for ordinal parameters, and x2 test for categorical parameters were used to compare baseline characteristics between groups. For measures of dependent parameters, a repeated-measures ANOVA was used to compare UFMA and JTT scores among time points (T0, T1, and T2).
For changes in OxyHb concentration, statistical parametric mapping (SPM) t-statistic maps were computed for group analyses and were considered significant at an uncorrected threshold of p < 0.05. Means, SDs, and 95% CIs were provided to depict the change within each group during the study and the training effect.