The Effect of an Augmented Reality-Based Mirror Therapy System on Upper Extremity Motor Recovery and Functional Activity Level in Stroke Patients

The aim of this study is to develop an augmented reality-based mirror therapy system, MirrARm, based on sensor-based tracking and visual reflection principles, and to investigate its effects on upper extremity motor recovery and functional activity levels in patients with stroke.

In this study, individuals with stroke who meet the inclusion criteria will be randomly assigned into two groups using stratification based on age, sex, and Fugl-Meyer Upper Extremity scores. The groups will be the Augmented Reality-Based Mirror Therapy group (AR-MT) and the Conventional Mirror Therapy group (C-MT). All participants will receive conventional physiotherapy training for 30 minutes per session, three days per week, for eight weeks. Participants in the AR-MT group will additionally receive 30 minutes of exercise training using the MirrARm system three days per week for eight weeks, while participants in the C-MT group will receive exercise training using conventional mirror therapy for the same duration and frequency.

In this study, augmented reality-based mirror therapy training will be delivered using the MirrARm system. The MirrARm system will consist of a 27-inch computer monitor, a Leap Motion Controller motion-tracking sensor, and a laptop computer. The monitor will be placed on a work table and will display images transmitted from the connected laptop computer. The Leap Motion sensor will also be connected to the laptop computer and positioned on the table in front of the monitor. The Leap Motion sensor will be used to transfer participants' hand movements to the screen in real time. Through its software features, the MirrARm system will display movements performed with one hand on the screen as if they were being performed by the opposite hand.

The software infrastructure of the MirrARm system will be developed using the Unity 6.1 game engine. Participants' hand movements will be tracked with high precision in a three-dimensional environment using the official software development kits (SDKs) and libraries of the Leap Motion device. Applications will be designed using ready-made or customized Unity assets positioned within a three-dimensional coordinate system in the Unity environment, and realistic physical interactions will be provided through Unity's NVIDIA PhysX-based physics engine. The operation of the application will be controlled through custom MonoBehaviour-based scripts developed using the C# programming language. These scripts will dynamically manipulate Transform component properties of scene objects, including position, rotation, and scale; process user interactions through event-based triggers; and execute predefined functions within the game loop. Objects requiring physical interaction will be assigned Unity's Rigidbody component to provide physical properties such as mass, gravity, and collision responses, and interaction regions will be defined using Collider components.

During training sessions using the MirrARm system, participants with stroke will be seated in front of the screen on a back-supported chair and will perform the tasks included in gamified therapeutic activities using their unaffected upper limb, while observing these movements on the screen as if they were being performed by the affected upper limb.

Four different activities to be used in the MirrARm system have been designed by the researchers based on movements considered important for upper extremity functionality after stroke. Each activity has been designed with three levels of difficulty. Participants will perform each level for two minutes, with a 30-second rest period between levels. The task-oriented activities of the MirrARm system are as follows:

• Cube stacking activity (shoulder elevation-focused)
• Reaching activity (elbow extension-focused)
• Supination-pronation activity
• Horizontal ring activity (wrist extension-focused)

All participants included in the study will be assessed three times during the intervention period (Week 0, Week 4, and Week 8) using primary and secondary outcome measures by a blinded assessor.

All statistical analyses of the data obtained from participants will be performed using SPSS version 27.0 (IBM Corp., Armonk, NY, USA). Continuous variables will be reported as mean ± standard deviation when the assumption of normal distribution is met and as median (interquartile range) when it is not. Categorical variables will be reported as number (N) and percentage (%). Assumption checks will include the Shapiro-Wilk test for normality, Levene's test for homogeneity of variances, and Mauchly's test for sphericity; in cases of sphericity violation, the Greenhouse-Geisser correction will be applied. When assumptions are met, a two-way mixed-design (2×3) analysis of variance (ANOVA) will be conducted for each outcome measure. The main effects of time and group, as well as the time × group interaction, will be tested, and multiple comparisons will be performed using Bonferroni correction. Effect sizes will be reported as partial eta squared (η²p), with 95% confidence intervals and a significance level set at p = 0.05. When assumptions are not met, within-group changes across three time points will be analyzed using the Friedman test, followed by Bonferroni-corrected pairwise Wilcoxon tests if significant. Between-group comparisons will be performed using the Mann-Whitney U test based on change scores across time points. Nonparametric results will be reported with effect sizes, 95% confidence intervals, and a significance level of p = 0.05.