Tracking Systems and Visualization Devices in Virtual, Augmented, and Mixed Reality Games for Motor and Cognitive Rehabilitation and Training: A Scoping Review

Abstract

Background: Virtual, augmented, and mixed reality (or collectively extended reality, XR) serious games, combined with motion-tracking technologies, are increasingly used for motor and cognitive rehabilitation and training. As XR and tracking technologies advance, a systematic mapping of the related research area could offer relevant insights. Objectives: This review aims to map interactive XR serious games, using motion-tracking technologies for physical or cognitive rehabilitation or training, and describe intervention characteristics and evaluation methods. Eligibility Criteria: Eligible studies were English, peer-reviewed journal articles published between 2015 and October 2025, with more than three participants, using custom XR serious games for rehabilitation or training. Studies were excluded if they focused on technical aspects, passive XR, diagnostic evaluation, psychological therapies, minor participants, procedural training, or education. Charting Methods: Data were charted using a structured form capturing XR characteristics, hardware configurations, study characteristics, and evaluation methods. Results: 61 studies were included. Most employed non-immersive or fully immersive VR interventions, targeting physical upper-body rehabilitation, especially post-stroke and Parkinson’s disease. Usability, acceptability and user experience, and training effectiveness were commonly evaluated with positive outcomes. Conclusions: The findings highlight opportunities for research into augmented and mixed reality approaches, particularly for cognitive function, and use of XR-based interventions across broader populations.

1. Introduction

1.1. Rationale

Extended reality (XR) technologies are being increasingly adopted across healthcare domains, as reflected by the growing research activity, expanding range of applications, and emerging clinical and educational uses [1,2,3]. A variety of XR applications, with modalities ranging from augmented reality (AR) to virtual reality (VR), are used to create immersive and interactive environments that aim to support physical, cognitive, and motor–cognitive interventions for rehabilitation and training.

1.2. Related Reviews in XR-Based Rehabilitation and Training

Many recent reviews have investigated the use of XR applications within specific clinical contexts, including rehabilitation following total knee arthroplasty [4], rehabilitation after acquired brain injury with the use of XR and mixed reality (MR) interventions [5,6], lower-limb rehabilitation employing AR applications [7], and XR interventions for stroke rehabilitation [8,9] as well as immersive VR applications for upper-limb rehabilitation in stroke populations [10]. Additionally, other reviews have focused on immersive XR interventions targeting psychological outcomes in spinal cord injury rehabilitation [11], VR and AR applications and exergames for chronic musculoskeletal rehabilitation [12], and VR-based interventions for postoperative cruciate ligament reconstruction [13].

Other reviews have adopted a more general approach without focusing on specific populations and pathologies. For example, a recent scoping review on AR and MR applications for motor rehabilitation mapped common sensing technologies, display devices, usability, and acceptability outcomes across a wide range of applications [14]. Similarly, ref. [15] mapped AR and MR applications supporting physical exercises in medical rehabilitation and proposed a taxonomy for patient types, rehabilitation purposes, and interaction design characteristics. Finally, a recent systematic review examined the effectiveness of integrating 3D and 4D digital human modeling within XR-based rehabilitation applications [16].

The existing reviews of XR interventions in health and rehabilitation highlight the breadth of this field and the wide variety of different applications and methodological approaches. This scoping review differs from previous population and pathology-specific reviews by adopting a broader perspective, not limited to a particular clinical condition. For example, while review [14] focused specifically on AR/MR applications for motor rehabilitation, the present review provides a system-oriented mapping of XR-based rehabilitation and training interventions across conditions, with a focus on motion-tracking technologies. Furthermore, this review includes all XR modalities (VR, AR, and MR), examines both motor and cognitive domains, and applies distinct eligibility criteria, including the exclusion of conference papers, studies with very small sample sizes, technically focused papers, single-session interventions, and interventions based on unmodified commercial off-the-shelf games. This allows the review to focus on structured, interactive XR interventions designed for rehabilitation or training rather than on technological systems alone. In summary, this scoping review aims to map XR-based rehabilitation and training interventions by examining the XR modalities, visualization devices, and motion-tracking technologies used, and to provide an overview of the different study characteristics and evaluation domains. This scoping review may provide a basis for a future systematic review.

1.3. Objectives

This scoping review aimed to map and summarize the existing literature of XR technologies used for physical and cognitive rehabilitation and training. More specifically, the review aimed to (i) identify XR system configurations and modalities, levels of immersion, visualization devices, and motion-tracking technologies used in the included studies, (ii) describe temporal trends in XR system configurations for the reviewed period, (iii) summarize the study characteristics of the included studies, and (iv) document the evaluation domains and measures used to assess XR-based rehabilitation and training interventions.

2. Methods

2.1. Protocol and Registration

No review protocol was registered for this review. To ensure the methodological transparency of this review, the PRISMA-ScR guidelines were followed [17].

2.2. Eligibility Criteria

To be eligible for inclusion in this review, studies needed to be peer-reviewed journal articles written in English. Other types of publications, such as reviews, perspective articles, or editorials, were excluded. While conference proceedings may be influential in XR and tracking research, they often report preliminary findings and are subjected to space constraints that may limit the level of methodological and technical detail provided. Also, the exclusion of conference proceedings reduced the likelihood of duplicate inclusion of extended journal versions of them. As the aim of this review was to systematically extract detailed information for system configurations, tracking technologies, intervention protocols, and outcome measures, inclusion was restricted to full-length peer-reviewed journal articles. This approach allowed only studies with more complete reporting to be included. Finally, only studies published after 2015 were considered, in order to ensure relevance to current technological developments of XR systems and motion-tracking methods.

Furthermore, to be included, a study should meet all the following criteria:

• Included three or more participants who completed a training or rehabilitation intervention. A minimum sample size of three participants was required to exclude single-case reports and very small-scale evaluations. This low threshold was selected to avoid excluding small pilot studies, which are common in emerging XR-based rehabilitation and training research.
• The intervention aimed to improve cognitive, physical, or motor–cognitive (physical and cognitive combined) abilities.
• The intervention was multi-session, conducted on separate days or visits.
• The intervention used a VR, AR, or MR application. This includes systems that are fully immersive (e.g., VR HMD), semi-immersive (e.g., CAVE, MR headset), or non-immersive (e.g., TV, screen, monitor).
• The application had to include at least one explicit gamification mechanic (e.g., defined goals or objectives, scoring systems, rewards such as stars, points or achievements, or progression elements). Alternatively, the intervention had to be explicitly described as a “game”, “serious game”, or “exergame”, with game elements verifiable from the text, figures, or supplementary materials. Supplementary videos or images were evaluated when available.
• The application had to be custom-developed for research purposes or a commercially available rehabilitation or training game developed by the study authors. Interventions based on commercially available, off-the-shelf games that could be purchased and used by the general public were excluded.

Studies were excluded if any of the following criteria were applicable:

• Included two or fewer participants.
• Participants did not take part in a training or rehabilitation intervention.
• The intervention was a single session.
• The intervention was based on passive experience (e.g., 360° videos).
• Focused on technical aspects (e.g., hardware, algorithms).
• The application functioned only as a diagnostic tool.
• The application could not be classified as a game (e.g., lack of gamification features).
• Used commercially available, off-the-shelf games without substantial modification (commercially available rehabilitation or training games were not excluded if the authors were part of the development team).
• Focused on procedural or professional skill training (e.g., surgery).
• Focused on knowledge transfer (educational applications).
• Focused on psychological therapy or psychological exposure applications.
• Included children or adolescent participants.

To enhance reproducibility and illustrate how eligibility criteria #5 and #6 were applied, screening followed a structured decision sequence, as described in Supplementary File S2.

2.3. Information Sources

To identify relevant studies, a comprehensive literature search was conducted in October 2025 using the Scopus, PubMed, IEEE Xplore, and ACM Digital Library databases. The results from each database were exported into Zotero 7.0 (Corporation for Digital Scholarship, VA, USA), where duplicate entries were identified and removed. One additional relevant study was identified through expert suggestion and included after assessment.

2.4. Search Strategy

The literature search was conducted in Scopus, PubMed, IEEE Xplore, and ACM Digital Library. The search was completed using a combination of terms defining the type of the immersive technology: “Virtual Reality”, “Augmented Reality”, “Mixed Reality”, “Extended Reality”, “VR”, “AR”, “MR”, and “XR” together with terms defining the type of motion tracking/recognition/sensing technology: “Motion Tracking”, “Motion Capture”, “MoCap”, “Body Tracking”, “Pose Estimation”, “Computer Vision”, “Object Tracking”, “Gesture Recognition”, “Skeleton Tracking”, “MediaPipe”, “OpenCV”, “Kinect”, “Head Mounted Display”, “HMD”, “Depth Camera”, “Human Pose Estimation”, “Human Motion Analysis”, “Hand Tracking”, “Face Tracking”, “Markerless Tracking”, “Depth Sensing”, “SLAM”, “Simultaneous Localization and Mapping”, “IMU”, “Visual-Inertial Odometry”, “VIO”, “Inertial Tracking”, “ARCore”, “ARKit”, and “Sensor Fusion”. The use of terms like “Head Mounted Display” or “HMD” was used because identifying a known HMD makes the sensing technology or method easy to determine.

Furthermore, the terms about motor or cognitive training included the following: “Cognitive Training”, “Physical Training”, “Exercise”, “Body Training”, “Rehabilitation”, “Dual Task Training”, and “Balance”.

Finally, the terms to describe a game-centered environment included the following: “Exergame”, “Exergaming”, “Serious Game”, “Virtual World”, “Gamification”, “Game”, “3D World”, “Virtual Environment”, “Immersive Experience”, “Interactive Simulation”, “Interactive Environment”, and “Interactive Experience”.

The terms within each of the main concept groups (type of immersive technology, type of sensing technology, type of training, and game-centered/interactive environment) were combined with OR, while the four groups were combined with AND. The final search structure followed the following pattern:

(Immersive technology terms) AND (Sensing technology terms) AND (type of training terms) AND (game-centered/interactive environment terms).

The above search terms were limited to “Title”, “Abstract”, and “Keywords” (or “metadata”, depending on the available database options). The range of publication years was limited to between January 2015 and October 2025. Depending on the available database filters, the results were also limited by document type to Articles, by type to Journals, and by language to English.

The full executable search strings for each database, including field specifications, applied filters, and record counts, are provided in Supplementary Material File S3.

2.5. Study Selection

The exported records from each database were imported into Zotero to identify and remove duplicates after manual verification. The remaining records were examined in two stages. First, the titles and abstracts were examined to exclude irrelevant publications, such as reviews and study protocols, which were identified from the titles and confirmed for exclusion using the abstracts. For the rest of the potentially relevant publications, the full-text articles were retrieved and assessed according to the eligibility criteria. Records for which the full text could not be retrieved were excluded. Both stages were assessed independently by two reviewers according to the eligibility criteria. Screening decisions were compared after each stage. Ambiguous cases were revisited to ensure the consistent application of the eligibility criteria, and disagreements were resolved through discussion. The percentage agreement between the two reviewers was calculated after screening and was 91% for title and abstract screening and 94% for full-text assessment.

2.6. Data Charting Process

2.6.1. General Data Extraction Procedure

The records from the selected databases were merged and organized before extraction. The data from the included studies were charted using Microsoft Excel 365 (Microsoft Corporation, Redmond, WA, USA), following a structured extraction template. The extracted information for the studies included bibliographic attributes and four main categories: the application characteristics, the technical and tracking specifications, the study details, and the evaluation outcomes.

The data charting template was refined during the initial stages of the review to ensure better capture of the relevant information and improve consistency. Brief notes were maintained for the excluded records. After the refinement, the final extraction template was used for all studies. Ambiguous cases were revisited to minimize subjective interpretation and maintain consistency.

2.6.2. Hardware Identification and Inference Rules

For the hardware, device models, tracking type, and any additional components were recorded when explicitly reported. When not specified, inference rules were applied. For studies not mentioning device version for “Kinect”, or whether it was Kinect for Xbox 360, Kinect for Windows, Kinect for Xbox One, or Kinect 2 for Windows (Microsoft Corporation, Redmond, WA, USA), the device version was inferred if clear supporting evidence was available. More specifically, if figures showed the Kinect 2 for Windows or Kinect for Xbox One form factor, the system was categorized as “Kinect v2”, while if images showed the Kinect for Xbox 360 or the characteristic silhouette of Kinect v1, then it was categorized as “Kinect v1”. In cases where hardware version identification was uncertain, more generic terms were used (e.g., “Kinect” for hardware and “RGB-D camera (markerless)” for the associated tracking method). This was required in a small number of cases (3 studies, about 5% of the included sample). Such classifications appear only in descriptive tables and do not affect the reported distribution patterns, as the analysis is based on higher-level categories such as modality, immersion level, visualization device type, and tracking technology type.

Low-level sensor components were not independently extracted unless explicitly described. Reported tracking systems, used only for the evaluation or validation of the performance or accuracy of the main tracking system, were not recorded as tracking hardware unless they were used directly for tracking related to the game interaction. Non–motion-tracking sensors (e.g., electromyography (EMG) and electroencephalography (EEG)) were excluded unless directly used for motion tracking or game interaction.

2.6.3. Terminology Normalization and Tracking Architecture Classification

Terms related to known commercially available tracking systems were normalized for consistency. For example, “Base Stations” in studies utilizing HTC Vive (HTC Corporation, Taipei, Taiwan) systems were standardized as “Lighthouse Base Stations”. When studies built around HTC Vive systems depicted or described worn trackers, then “Vive Trackers” were recorded. Normalization of hardware terminology was limited to known commercial systems and only based on text or visual evidence.

For the classification of the tracking architecture of head-mounted displays (HMDs) (optical inside-out and outside-in), the technical specifications of the identified commercial system were used. For example, Meta Quest 2 (Meta Platforms Inc., Menlo Park, CA, USA) and Oculus Rift S (Oculus VR, LLC, Menlo Park, CA, USA) systems were categorized as optical inside-out, whereas HTC Vive Pro (HTC Corporation, Taipei, Taiwan) systems that use Lighthouse base stations were categorized as outside-in. Tracking types were not inferred beyond the technical characteristics of commercially available hardware.

2.6.4. Harmonization of Body Domains and Clinical Conditions

To improve comparability across the included studies, body areas like the knee or lower limbs and related functions such as gait were grouped under “lower body”, while areas like the trunk, upper extremity, shoulder, or neck were grouped under “upper body”. Similarly, “full body” categorized studies that either explicitly targeted the full body or targeted areas of both the upper and lower body.

Studies reporting different subtypes of the same condition were grouped into the broader condition category. For example, stroke with chronic upper extremity hemiparesis, chronic stroke, subacute stroke, or just stroke were all aggregated under the category “stroke”. This aggregation was performed to allow comparison across studies.

2.7. Data Items

The extracted data from the studies covered bibliographic attributes (title, first author, publication year, and DOI) and variables grouped in four categories: application characteristics, technical and tracking specifications, study and participant characteristics, and evaluation outcomes.

2.7.1. Application Characteristics

Application characteristics included the type of XR modality (VR, AR, and MR), level of immersiveness, game theme, and development platform.

The XR modality was recorded according to the terminology used by the study. When the terminology was inconsistent, classification was based on the capabilities of the visualization device described. Game themes were summarized according to the authors’ descriptions.

2.7.2. Technical and Tracking Specifications

Technical specifications included the visualization device, tracking hardware, and tracking type.

When hardware was not explicitly described, or details such as version were missing, contextual details or device images and illustrations within the study were used to infer the device variant (e.g., Kinect v1 vs. v2). The tracking type was based on the motion tracking hardware (e.g., “RGB-D camera (markerless)” for Kinect, or “inside-out” for Meta Quest 2). The term “Hybrid tracking” was used to classify studies that utilized more than one major motion-tracking approach within the same intervention, without necessarily implying sensor fusion.

2.7.3. Study and Participant Characteristics

This category included the target body domain, intended application (rehabilitation or training), target function (physical, cognitive, or both), target population/condition, sample size, whether the sample matched the intended population, and the intervention duration.

The target body domain was harmonized into three categories (upper body, lower body, full body) and was based on the primary study focus. If body domains were not explicitly stated, the evaluation tools employed by the study were used to determine the relevant body domain. Pathology labels were standardized where possible (e.g., “subacute stroke” was categorized as “stroke”) to enhance cross-study comparability.

2.7.4. Evaluation Outcome Categories

Evaluation outcomes were grouped into six predefined constructs: usability; acceptability and user experience (UX); training effectiveness; cybersickness; system accuracy; and measurement validity. The definitions and coding criteria were defined in advance for consistent classification.

Usability

Usability was defined according to ISO 9241-11 [18] as the extent to which a system enables users to achieve their goals with effectiveness, efficiency, and satisfaction for a specified context of use.

Standardized or validated usability tools, such as the System Usability Scale (SUS) [19] and the User Satisfaction Evaluation Questionnaire (USEQ) [20], were categorized as usability measures. UX evaluation tools, such as the User Experience Questionnaire (UEQ) [21], were categorized as usability when reported outcomes reflected efficiency or ease of use.

Instruments designed for assessing efficiency, ease of use, learnability, or task-related workload, such as the Raw Task Load Index (RTLX) and NASA Task Load Index (NASA-TLX) [22], were categorized under usability. The Quebec User Evaluation of Satisfaction (QUEST) [23] was also included in the usability category.

Non-standard evaluation tools, such as qualitative feedback or custom questionnaires, were categorized as usability measures if they aimed to assess usability-related items, such as effectiveness or efficiency of use, ease of use, or learnability.

Acceptability and User Experience

Acceptability and UX were categorized under the same construct. Acceptability according to the Technology Acceptance Model (TAM) [24,25] can be primarily described as perceived usefulness and perceived ease of use, which influences the attitudes and behavior of a user toward using and the actual system usage.

Standardized or validated tools designed to assess technology acceptance, such as the Unified Theory of Acceptance and Use of Technology (UTAUT) [26] and the Technology Usage Inventory (TUI) [27], were categorized under this construct. Questionnaires (e.g., Physical Exercise Adherence Questionnaire and questionnaires for overall experience) and evaluation measures that aimed to assess acceptability-related items such as adherence, motivation, engagement, compliance, intention to use, or usage frequency were also included.

UX-related tools and questionnaires designed to measure constructs such as enjoyment, affect, attractiveness, or immersion, which may influence technology acceptance, were categorized within this construct. Examples of UX tools included the Intrinsic Motivation Inventory (IMI), the Game Experience Questionnaire (GEQ), the Positive and Negative Affect Schedule (PANAS), the Self-Assessment Manikin (SAM), the Perceived Restoration Scales (PRS), and the Presence in AR questionnaire (pAR).

When referring to the GEQ, a distinction was made between the UX related GEQ [28] and usability satisfaction measures derived from the IBM Computer Usability Satisfaction Questionnaires [29], which were categorized under usability.

Training Effectiveness

Training effectiveness referred to participant improvement or functional progress that was a result of the intervention. This construct included standardized clinical assessments and performance-based functional tests, such as the Fugl–Meyer Assessment (FMA), Berg Balance Scale (BBS), Timed Up and Go (TUG), Montreal Cognitive Assessment (MoCA), and Wolf Motor Function Test (WMFT). It also included game-derived performance metrics used to evaluate improvement, as well as pre–post comparisons and between-group analyses of physical or cognitive outcomes.

Cybersickness

Cybersickness refers to sickness-like symptoms such as eye strain, headache, nausea, or disorientation that are related to real-time 3D graphics, especially when using HMDs [30]. Formal assessment included use of standardized or validated tools such as the Simulator Sickness Questionnaire (SSQ) [31] and the Virtual Reality Sickness Questionnaire (VRSQ) [32]. Other assessment methods, including custom questionnaires, checklists, or qualitative feedback aimed at capturing motion sickness symptoms or adverse events and effects, were also included.

System Accuracy

System accuracy referred to the performance, precision, and reliability of the motion-tracking equipment and sensors used for the intervention. Evaluation measures included assessment of tracking precision and accuracy, agreement between sensor outputs and external measurement tools, repeatability, error rates, or internal system-algorithm evaluations (e.g., posture-matching or gesture detection accuracy).

Measurement Validity

Measurement validity referred to the extent to which system metrics reflected the functions they were intended to measure. Measurement validity was considered assessed when studies examined the relationships between system metrics and external measures, including correlation or regression analyses, or comparisons with established external measures.

2.7.5. Data Charting and Classification Procedures

The data extraction and classification of the evaluation measures into the predefined constructs were performed independently by two reviewers. The coding decisions were compared, and disagreements were resolved through discussion until agreement on the classifications was reached. When the construct boundaries were ambiguous (e.g., usability versus UX), classification was guided by the study descriptions and the theoretical foundation of the used instruments when commonly used tools were involved.

Illustrative examples of evaluation measure classification from ten representative included studies are provided in Supplementary File S4.

2.7.6. Outcome Categorization

Evaluation outcomes and their descriptive records were based on the reported findings of each study. The reported findings were categorized as favorable when the study reported improvements or positive results for the primary evaluation measures according to the authors of that study. When widely adopted standardized tools were used (e.g., SUS), categorization followed the scoring guidance of the tool. No independent reassessment or reanalysis of statistical results was conducted.

2.8. Critical Appraisal of Individual Sources of Evidence

Study quality evaluation or risk of bias assessment was not performed, as the aim of this scoping review was to map the existing evidence.

2.9. Presentation of Results

The extracted data were synthesized descriptively and organized using the data-charting categories. Findings were grouped according to the objectives. Tables and narrative summaries were used to present them and were accompanied by visual representations (e.g., bar charts and diagrams) when appropriate. Additionally, when possible, terminology inconsistencies were normalized to improve uniformity of the reported items and consistent grouping across the different studies. No meta-analysis was conducted, as the aim of this scoping review was to map the existing evidence.

3. Results

3.1. Selection of Sources of Evidence

From a total of 752 records identified from the four databases after applying all available filters (e.g., publication type and language), 170 were removed as duplicates. The remaining 582 records were screened by title and abstract to remove studies that could be readily identified as reviews, study protocols, or involving ineligible populations, and 58 studies were removed. Full text was unavailable for 22 records. The full texts of the remaining 502 studies were assessed for eligibility. Of these, 442 were excluded for the following main reasons: 14 were reviews, 1 was a study protocol, 3 were perspective articles, 1 was a book chapter, and 1 was a conference article. A total of 78 studies had a primarily technical focus, included no participants, or were unrelated to rehabilitation or training. A total of 194 studies did not meet rehabilitation or training criteria (e.g., single-session or single-visit studies, no reported intervention duration, or diagnostic focus). A total of 11 studies used non-interactive content (e.g., 360° videos). A total of 14 involved ineligible populations. A total of 24 studies were excluded for not being games or lacking gamification features. A total of 14 studies focused on procedural skills or education and were excluded. A total of 79 used unmodified commercial games or platforms. Five were case studies or had two or fewer participants. One article was not in English, and two were alternative versions of previously assessed studies. One additional study identified through other sources (expert suggestion) was retrieved and assessed for eligibility. In total, 61 studies (60 identified through database searching and 1 from other sources) were included in the scoping review. The study selection process is illustrated according to the PRISMA flow diagram in Figure 1.

Figure 1. PRISMA flow diagram.

3.2. Characteristics of Sources of Evidence

All 61 included articles were published between 2015 and 2025, were multi-session interventions, and focused on interactive serious games or exergames for physical, cognitive, or motor–cognitive rehabilitation or training.

Most of the described serious games employed VR systems, while only a small number of studies employed AR or MR platforms. The immersiveness of the covered systems ranged from non-immersive to fully immersive. Non-immersive games were most common, especially among earlier studies, while fully immersive systems that use HMDs were more prominent in later studies. Semi-immersive MR systems were less frequent. Some of the studies described platforms utilizing more than one level of immersiveness. Visualization devices included screen-based systems, VR HMDs, and MR headsets.

Motion tracking was implemented using a wide variety of technologies. Optical markerless approaches relying on RGB-D cameras were the most commonly used. Other, non-optical-based motion tracking methods described were inertial and hybrid configurations.

Most studies focused on rehabilitation serious games, while a smaller number focused on training applications for healthy participants or populations at risk. All the included studies involved adult participants, with some of them targeting healthy populations while others focused on clinical populations with conditions such as stroke or Parkinson’s disease. Sample sizes were generally small to moderate, and the duration of the interventions ranged from a few sessions over a single week to programs extending up to six months.

Training effectiveness was the most reported evaluation outcome, followed by acceptability and UX and usability. System accuracy, metrics validity, and cybersickness were less frequently reported outcomes.

3.3. Critical Appraisal Within Sources of Evidence

There was no critical appraisal of the included sources of evidence conducted, as this was outside the scope of the review.

3.4. Results of Individual Sources of Evidence

Table 1 presents an overview of the included sources of evidence with the serious-game activities or tasks implemented in each intervention. Table 2 reports the XR system characteristics, including XR modality, immersiveness, development platform or game engine, visualization device, tracking hardware, and primary tracking type employed. Table 3 presents the study characteristics, such as the participant characteristics, the duration, and the body domain targeted by the intervention. Table 4 and Table 5 present the evaluation domains (usability, acceptability and UX, training effectiveness, cybersickness, system accuracy, and metrics validity), and the related outcomes.

Table 1. Primary objective and game activities/tasks of the included sources of evidence.

Study	Year	Primary Objective	Game Activities/Tasks
[33]	2015	To examine whether motor performance metrics derived during virtual activities of daily living reflect upper-extremity motor function in stroke patients	VOTA: daily-life activities (e.g., Meal preparation)
[34]	2018	To evaluate the effectiveness of virtual activities of daily living practice for improving upper-extremity motor function in chronic stroke survivors	SaeboVR: 12 ADL activities (e.g., Grocery Shopping; Putting Away Groceries; Preparing Breakfast; Pet Shopping; Pet Feeding; Pet Bathing; Garden Planting)
[35]	2020	To evaluate the usability and therapeutic effectiveness of an adaptive XR-based serious game for telerehabilitation of patients with chronic neck pain	Rehabot: Runner-style rehabilitation exercises
[36]	2022	To assess the feasibility and preliminary effects of a home-based MR mirror therapy system for managing phantom limb pain	Mr. MAPP: Bubble Burst; Pedal; Piano
[37]	2018	To evaluate the effects of Kinect-based VR game training on upper-extremity motor recovery in patients with chronic stroke	KineLabs: Good View Hunting; Hong Kong Chef (Egg Tart; BBQ Pork with Rice)
[38]	2019	To present and evaluate a low-cost immersive VR serious-games framework for full-body motor rehabilitation with objective performance assessment	Plank Walk; Single-stance; Finger exercise
[39]	2018	To develop and evaluate a fully immersive VR rehabilitation system that uses a Long Short-Term Memory Recurrent Neural Network (LSTM-RNN) to assess patient motor performance	VRheab: Raise knee and pinch fingers; March and grasp; Get up on heels or toes and cut spheres
[40]	2020	To develop and evaluate an immersive VR platform for neurocognitive rehabilitation using daily tasks	Spatial navigation tasks
[41]	2023	To evaluate the effects of a Kinect-based rehabilitation training software on mobility in patients with Parkinsonian syndromes	Exercises: Coconuts; Stars; Balls; Balloons
[42]	2021	To investigate the effects of balance-based exergaming training on biomechanical measures of balance in patients with Parkinson’s disease	Boat; Colors; Froglet; Football player; Burro; Bicycles; Fruits
[43]	2022	To evaluate the feasibility, safety, acceptability, and potential cognitive benefits of immersive VR-based cognitive rehabilitation in stroke survivors	VIRTUE ADL activities (e.g., Make bed; Select clothes; Brush teeth; Make toast; Order meal; Water plants)
[44]	2016	To develop and evaluate a Kinect-based interactive exercise system to improve physiological function and standing stability in older adults	Virtual soccer goalkeeper
[45]	2016	To develop and evaluate a Kinect-based virtual training system for shoulder rehabilitation and performance monitoring	Shoulder finger ladder; Single curved shoulder
[46]	2021	To co-design and evaluate a bowling exergame for older adults and compare functional capacity and exercise adherence when playing alone versus with peers	Boliche Virtual (Bowling exergame)
[47]	2023	To evaluate the usability, feasibility, and functionality of a robotic upper-limb rehabilitation system combined with an AR serious game	RobExReha: 3D puzzle
[48]	2025	To develop and evaluate a user-centered VR-based adaptive mirror therapy system for upper-limb post-stroke rehabilitation	VVITA: Reach and grasp tasks (Beach ball; Concertina)
[49]	2022	To examine whether an immersive VR serious game can sustain engagement and support upper-extremity physical rehabilitation beyond the novelty effect	Project Butterfly (Protecting a virtual butterfly)
[50]	2018	To develop and evaluate a non-immersive VR rehabilitation game for assessing and supporting upper-limb motor function in individuals with neurological impairment	ReHabgame: Reach-Grasp-Release fruits; Reach-Press-Hold buttons
[51]	2022	To evaluate the cost-effectiveness of a VR game–based back extension program compared with clinic-based McKenzie therapy for chronic non-specific low back pain	BE-VRG: Back Extension–VR Game (reach for virtual balls)
[52]	2023	To assess the feasibility and acceptance of AR and VR exergames for training motor and cognitive skills in older adults	GAME2AWE: Fun Park Tour (Whack-a-Mole; Bazaar; Darts); Life on a Farm (Farm activities)
[53]	2023	To examine the feasibility and effects of the COGNIPLAT serious game platform for multi-domain cognitive training in older adults with mild cognitive impairment	COGNIPLAT: Multi-domain cognitive training tasks
[54]	2025	To assess the feasibility and preliminary motor–cognitive effects of a multi-technology exergaming intervention for older adults	GAME2AWE: Fun Park Tour (Whack-a-Mole; Bazaar; Darts); Farm Living (Farm activities)
[55]	2024	To evaluate the feasibility and potential effects of a remotely prescribed, monitored, and tailored home-based AR gait-and-balance exergaming intervention for people with Parkinson’s disease	Reality DTx®: Gait and balance training games (Mole Patrolll; Smash!; Hot Buttons; Basketball; Walking)
[56]	2025	To examine the feasibility and effects of a VR blowgun game on swallowing and respiratory functions in older adults	VR Blowgun dartboard game
[57]	2022	To compare the effects of supervised and non-supervised upper-limb VR exercises on sensory–motor function in patients with idiopathic Parkinson’s disease	ULVRE: Main reach; Random reach; Tracking; Rotating wheel; Library
[58]	2022	To examine whether specific game design principles improve engagement of stroke survivors using a desktop VR serious game for upper-limb rehabilitation	Neuromender Rehabilitation System: The Wingman (Position toward rings)
[59]	2015	To design and evaluate a low-cost robotic exergame system using the unaffected arm to support upper-limb rehabilitation in stroke patients	Moving a cup task
[60]	2017	To evaluate the effects of a VR-based fine hand motion rehabilitation system on motor recovery in subacute stroke patients	Flying Bird; Spaceship; Supermarket; Kitchen and cooking; SpaceWar
[61]	2024	To develop and evaluate a glove-based virtual hand rehabilitation system and assess its effects on range of motion and grip strength in patients with post-traumatic hand injuries	RehabRelive Glove: Runman; Squeezy
[62]	2018	To assess the feasibility and preliminary effects of a serious games platform for upper-limb rehabilitation in persons with multiple sclerosis	Rehab@Home (Six arm rehab games)
[63]	2020	To investigate the feasibility, usability, and preliminary effects of a fully immersive VR upper-extremity rehabilitation program using a head-mounted display in stroke patients	Hammering; Catch ball; Cup pouring; Bubble touching; Xylophone playing
[64]	2018	To evaluate the effects of an attention-based VR upper-limb rehabilitation system on motor recovery in post-stroke patients	Cannon shooting game
[65]	2018	To design and preliminarily evaluate a therapist-driven system for creating patient-oriented rehabilitation video games using Kinect-based motion tracking	Reach the Apple; Keep the Ball
[66]	2023	To evaluate the effects of a gamified VR graded-activity intervention on pain and related outcomes in people with chronic low back pain	RoBow Agent
[67]	2022	To evaluate the effects of an intradialytic non-immersive VR exercise program on physical function and adherence in patients undergoing haemodialysis	Treasure Hunt (Catch targets; Avoid obstacles)
[68]	2018	To assess the feasibility, acceptability, and effects of a customized videogame rehabilitation program on freezing of gait and balance disorders in patients with Parkinson’s disease	Toap Run (Collect coins; Avoid obstacles)
[69]	2024	To evaluate the efficacy, safety, and acceptability of a home-based tailored exergaming intervention for gait and balance disorders in patients with Parkinson’s disease	Toap Run (Collect coins; Avoid obstacles)
[70]	2015	To develop and evaluate a Kinect-based game system for lower-limb rehabilitation aimed at improving gait performance in patients with Parkinson’s disease	Whack-a-Mole style leg exercise
[71]	2015	To examine the feasibility and utility of a customized, in-home Kinect-based game intervention for individuals with chronic stroke	Mystic Isle (ADL activities)
[72]	2023	To validate the performance effects of a VR squat exergame with force-feedback and dynamic difficulty adjustment on lower-limb muscle activity and squat execution	Ski course (Squat to collect objects)
[73]	2024	To evaluate whether an affect-adaptive VR environment can increase active muscle effort during assisted gait rehabilitation	VR MoonWalker
[74]	2015	To examine the feasibility of using a virtual environment with natural gesture-based interaction to teach street-crossing and path-following safety skills to adults with autism spectrum disorder	Street-crossing; path-following
[75]	2024	To evaluate the technical feasibility and initial performance of a gamified computer-vision–based platform for balance training and monitoring in patients with balance disorders	Reach or strike moving targets (e.g., Virtual balls interaction)
[76]	2025	To develop and preliminarily evaluate a high-fidelity immersive HMD-VR exergame with restorative virtual environments for gait and balance rehabilitation in older adults	Walking environments (Urban; Forest; Rural)
[77]	2024	To investigate the effects of a home-based, full-body exergaming intervention on gait, balance, and associated brain connectivity in people with advanced Parkinson’s disease	Toap Run (Collect coins; Avoid obstacles)
[78]	2023	To compare the effects of an immersive VR serious game based on activities of daily living with conventional computerized cognitive training in inpatient neurorehabilitation	Teora mind: Daily-life activities (Kitchen; Garden)
[79]	2022	To examine the preliminary effectiveness and UX of an immersive VR exergame as a supplement to multimodal pain therapy in older adults with chronic back pain	ViRST: Exercises activities (e.g., Rowing; Turning on light bulbs; Sorting vegetables)
[80]	2022	To compare the usability and acceptance of two simulated MR exergames representing strength endurance and endurance training in older adults with hypertension	Strength endurance training (Squat; Overhead press; Lateral raise; Leg raise; Heel raise); Endurance training (Ball game; High five; Hustle dance)
[81]	2016	To evaluate the usability, acceptance, and rehabilitation effects of a Kinect-based 3D game system for patients undergoing total knee replacement rehabilitation	Bike-racing
[82]	2016	To evaluate the effects of a Kinect-based gamified rehabilitation system on knee function, self-efficacy, and rehabilitation achievement after total knee replacement	Fruit kicking
[83]	2025	To evaluate the effects of immersive VR-assisted rehabilitation on neuromuscular function, pain, and psychosocial outcomes in patients with postoperative C5 palsy	Apple placing tasks
[84]	2020	To compare participation, motivation, and upper-extremity motor outcomes between multiuser and single-user home-based VR therapy after stroke	VERGE: Ball Bump; Food Fight; Trajectory Trace
[85]	2018	To develop and evaluate the feasibility and user response of a multi-user, networked VR system for home-based upper-limb rehabilitation after stroke	VERGE: Ball Bump; Food Fight; Trajectory Trace
[86]	2024	To investigate the effects of VR–based skateboarding training on balance, functional performance, muscle force, and pain in patients with degenerative lumbar spine disease	VR skateboarding training
[87]	2020	To develop and evaluate an HMD-based VR maze task for memory assessment and cognitive training	Maze task
[88]	2017	To present a Kinect-based system for home-based cardiovascular rehabilitation and collect user feedback	Kinect-RehabPlay exercise program
[89]	2023	To evaluate the effects of a home-based VR exercise program during the maintenance stage of cardiac rehabilitation on lower-limb strength, physical activity, and exercise tolerance	Kinect-RehabPlay exercise program
[90]	2020	To develop and evaluate a Kinect-based interactive game system for shoulder rehabilitation and objective motion assessment using multi-sensor data	Whitewater Rafting; Collision Pinball
[91]	2016	To evaluate feasibility and effects of home-based IMU-driven VR arm rehabilitation after stroke	ArmeoSenso: Meteors; Slingshot
[92]	2024	To evaluate the effects of a VR-based walker avoidance training program on step-aside reaction time and collision avoidance performance	Walker avoidance game
[93]	2021	To evaluate the usability and clinical effects of a depth camera–based, task-specific VR game for lower-extremity rehabilitation after stroke	Stomp Joy (Whack-A-Mole style game)

Table 2. XR hardware, software, and interaction characteristics.

Study	XR Modality	Immersiveness	Development Platform/Game Engine	Visualization Device	Tracking Hardware	Main Tracking Type
[33]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect v1	RGB-D camera (markerless)
[34]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect v2	RGB-D camera (markerless)
[35]	VR	Fully immersive; Non-immersive	Unity	HTC Vive; Screen/monitor	Kinect v2; HTC Vive	Hybrid: RGB-D camera (markerless); Optical outside-in
[36]	MR	Fully immersive	Mr. MAPP framework	Oculus Rift	Kinect v2	RGB-D camera (markerless)
[37]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect v1	RGB-D camera (markerless)
[38]	VR	Fully immersive	Unity	Oculus Rift DK2	Kinect v2; Oculus Rift DK2; Leap Motion Controller	Hybrid: RGB-D camera (markerless); Optical outside-in; Infrared optical (hand tracking)
[39]	VR	Fully immersive	Unity	Oculus Rift	Kinect v2; Oculus Rift; Leap Motion Controller	Hybrid: RGB-D camera (markerless); Optical outside-in; Infrared optical (hand tracking)
[40]	VR	Fully immersive	Unity	HTC Vive	HTC Vive; Lighthouse Base Stations; Controllers	Optical outside-in
[41]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect	RGB-D camera (markerless)
[42]	VR	Semi-immersive	Not reported	Large Screen/Monitor	Kinect v2; force platform	Hybrid: RGB-D camera (markerless); Force sensing
[43]	VR	Fully immersive	Unity	Oculus Rift S	Oculus Rift S; Touch Controllers	Optical inside-out
[44]	VR	Non-immersive	Unity	Screen/Monitor	Kinect v1	RGB-D camera (markerless)
[45]	VR	Non-immersive	Unity	Screen/Monitor	Kinect v1	RGB-D camera (markerless)
[46]	VR	Non-immersive	Unity	Screen/Monitor	Kinect v2	RGB-D camera (markerless)
[47]	AR	Semi-immersive	Unity	HoloLens	HoloLens; LBR iiwa	Hybrid: Optical inside-out; Mechanical/robotic
[48]	VR	Fully immersive	Not reported	HTC Vive Pro	HTC Vive Pro; Vive Trackers; Lighthouse Base Stations; Myo Armbands (EMG)	Hybrid: Optical outside-in; Inertial (IMU-based); EMG-based
[49]	VR	Fully immersive	Unity	HTC Vive Pro	HTC Vive Pro; Lighthouse Base Stations; Controllers	Optical outside-in
[50]	VR	Non-immersive	Unity	Screen/Monitor	Kinect v2; Myo Armband	Hybrid: RGB-D camera (markerless); Inertial (IMU-based); EMG-based
[51]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect	RGB-D camera (markerless)
[52]	VR; AR	Fully immersive (VR); Non-immersive (AR)	Unity; Unreal Engine	Oculus Quest 2; Screen/Monitor	Kinect v2; Oculus Quest 2; Controllers; Smart floor; Android (ARCore)	Hybrid: RGB-D camera (markerless); Optical inside-out; Force/pressure sensing
[53]	VR; AR	Non-immersive	Not reported	Screen/Monitor	Tablet (tablet camera), Touchscreen	Optical inside-out
[54]	VR	Fully immersive; Non-immersive	Not reported	Oculus Quest 2; Screen/Monitor	Kinect v2; Oculus Quest 2; Controllers; Smart floor	Hybrid: RGB-D camera (markerless); Optical inside-out; Force/pressure sensing
[55]	AR	Semi-immersive	Not reported	Magic Leap 2; HoloLens 2	HoloLens 2; Magic Leap 2	Optical inside-out
[56]	VR	Fully immersive	Not reported	Meta Quest 2	Meta Quest 2; microphone	Optical inside-out
[57]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect v1	RGB-D camera (markerless)
[58]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect	RGB-D camera (markerless)
[59]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect; FSR sensor Glove	Hybrid: RGB-D camera (markerless); Force/pressure sensing
[60]	VR	Fully immersive; Non-immersive	Unity	Oculus Rift DK2; Screen/Monitor	Oculus Rift DK2; Oculus position tracker; Amadeo robotic hand	Hybrid: Optical outside-in; Mechanical/robotic
[61]	VR	Non-immersive	Not reported	Screen/Monitor	RehabRelive Glove (12 IMU MPU9250 sensors)	Inertial (IMU-based)
[62]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect v2	RGB-D camera (markerless)
[63]	VR	Fully immersive	Not reported	HTC Vive	HTC Vive, controllers	Optical outside-in
[64]	VR	Non-immersive	Unity	Screen/Monitor	Custom Motion Capture Device (accelerometer; gyroscope; magnetometer; infrared sensor)	Inertial (IMU-based)
[65]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect v2	RGB-D camera (markerless)
[66]	VR	Fully immersive	Unity	Oculus Rift	Oculus Rift; Controllers	Optical outside-in
[67]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect v1	RGB-D camera (markerless)
[68]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect v2	RGB-D camera (markerless)
[69]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect v2	RGB-D camera (markerless)
[70]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect v1	RGB-D camera (markerless)
[71]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect v1	RGB-D camera (markerless)
[72]	VR	Fully immersive	Unity	HTC Vive	HTC Vive; Vive Trackers; Lighthouse Base Stations	Optical outside-in
[73]	VR	Fully immersive	Unity	Meta Quest 2	Meta Quest 2; controllers	Optical inside-out
[74]	VR	Semi-immersive	NeuroVR 2.0	Large projection screen	Kinect v1	RGB-D camera (markerless)
[75]	VR	Non-immersive	Not reported	Screen/Monitor	Intel RealSense D435i	RGB-D camera (markerless)
[76]	VR	Fully immersive	Unity	HTC Vive Pro Eye	HTC Vive Pro Eye; Lighthouse Base Stations	Optical outside-in
[77]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect v2	RGB-D camera (markerless)
[78]	VR	Fully immersive	Not reported	Oculus Quest	Oculus Quest, hand controllers	Optical inside-out
[79]	VR	Fully immersive	Not reported	HTC Vive	HTC Vive; Controllers; Vive Trackers; Lighthouse Base Stations	Optical outside-in
[80]	VR	Fully immersive	Not reported	HTC Vive Pro	HTC Vive Pro; Vive Trackers; Lighthouse Base Stations; Controllers	Optical outside-in
[81]	VR	Non-immersive	Unity	Screen/Monitor	Kinect v1	RGB-D camera (markerless)
[82]	VR	Non-immersive	Unity	Screen/Monitor	Kinect v1	RGB-D camera (markerless)
[83]	VR	Fully immersive	Not reported	HTC Vive Pro 1.0	HTC Vive Pro 1.0; Lighthouse Base Stations; Controllers	Optical outside-in
[84]	VR	Non-immersive	Unity	Screen/Monitor	Kinect v1	RGB-D camera (markerless)
[85]	VR	Non-immersive	Unity	Screen/Monitor	Kinect v1	RGB-D camera (markerless)
[86]	VR	Fully immersive	Unity	HTC Vive HMD	HTC Vive; Vive Trackers; Lighthouse Base Stations	Optical outside-in
[87]	VR	Fully immersive	Unity	Oculus Go	Oculus Go; controller	Inertial (IMU-based)
[88]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect v1	RGB-D camera (markerless)
[89]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect v1	RGB-D camera (markerless)
[90]	VR	Non-immersive	Not reported	Screen/Monitor	Kinect v1; Bluetooth ball (IMU)	Hybrid: RGB-D camera (markerless); Inertial (IMU-based)
[91]	VR	Non-immersive	Not reported	Screen/Monitor	MotionPod 3 (IMUs)	Inertial (IMU-based)
[92]	VR	Fully immersive	Unity	Meta Quest 2	Meta Quest 2	Optical inside-out
[93]	VR	Non-immersive	Not reported	Screen/Monitor	PrimeSense 3D Awareness Sensor	RGB-D camera (markerless)

Table 3. Study characteristics for the included interventions.

Study	Target Body Domain	Intended Application	Targeted Function	Target Population/Condition	Sample Size	Sample–Target Match	Intervention Dosage
[33]	Upper body	Rehabilitation	Physical	Stroke	N = 14	YES	2 weeks; 4 sessions; 1 h/session
[34]	Upper body	Rehabilitation	Physical	Stroke	N = 15	YES	8 weeks; 24 sessions; 3×/week; 1 h/session
[35]	Upper body	Rehabilitation	Physical	Chronic neck pain	N = 14	YES	2 weeks; 10 sessions; 5×/week; 10 min/session
[36]	Lower body	Rehabilitation	Physical	Phantom limb pain	N = 4	YES	1 month; 2× sessions/day; 1 h; home-based, self-directed
[37]	Upper body	Rehabilitation	Physical	Stroke	N = 38 (18 IG; 20 CG)	YES	4 weeks; 20 sessions; 5×/week; 1 h per session
[38]	Full body	Rehabilitation	Physical	Stroke	N = 92	YES	4 weeks
[39]	Full body	Rehabilitation	Physical	Parkinson’s disease	N = 5	YES	10 sessions; 45–60 min/session
[40]	Full body	Rehabilitation	Cognitive	Stroke	N = 35	YES	3 weeks; 9 sessions; 3×/week
[41]	Full body	Rehabilitation	Physical	Parkinson’s disease	N = 74 (25 IG1; 24 IG2; 25 CG)	YES	3 weeks; IG1: 2×/week, IG2: 5×/week; 20 min/session
[42]	Full body	Rehabilitation	Physical	Parkinson’s disease	N = 24 (12 IG; 12 CG)	YES	4 weeks; 12 sessions; 3×/week; 30 min/session
[43]	Upper body	Rehabilitation	Cognitive	Stroke	N = 40 (30 IG; 10 CG)	YES	≤2 weeks; 5×/week; inpatient
[44]	Full body	Training	Physical	Healthy older adults	N = 40 (20 G1; 20 G2)	YES	12 weeks; 2×/week; 30 min/session
[45]	Upper body	Rehabilitation	Physical	Shoulder impairment	N = 6	YES	6 weeks; 2×/week; 3 rounds/session
[46]	Full body	Training	Physical	Healthy older adults	N = 19 (10 with peers; 9 alone)	YES	10 weeks; 21 sessions; 2×/week; 10 min/session
[47]	Upper body	Rehabilitation	Physical	Stroke	N = 11	YES	4–5 sessions; 45 min/session
[48]	Upper body	Rehabilitation	Physical	Stroke	N = 4	YES	1 month; 3×/week; 30 min/session
[49]	Upper body	Rehabilitation	Physical	Upper-extremity injury	N = 5	YES	8 weeks; 2×/week; 45–60 min/session
[50]	Upper body	Rehabilitation	Physical	Stroke	N = 20	YES	8 weeks; 60 min/session
[51]	Full body	Rehabilitation	Physical	Chronic non-specific low back pain	N = 46 (22 IG; 24 CG)	YES	8 weeks; 24 sessions; 3×/week; 30 min/session
[52]	Full body	Training	Physical; Cognitive	Healthy older adults	N = 23	YES	5 weeks; 2×/week; 30 min/session; pilot
[53]	Upper body	Training	Cognitive	Mild cognitive impairment (MCI)	N = 21 (11 IG; 10 CG)	YES	12 weeks; 24 sessions; 1–2×/week; 60 min/session
[54]	Full body	Training	Physical; Cognitive	Older adults	N = 58 (30 IG; 28 CG)	YES	16 weeks; 1×/week; 60 min/session; 20 min active gameplay
[55]	Full body	Rehabilitation	Physical	Parkinson’s disease	N = 20	YES	6 weeks; 5×/week; 30 min/session
[56]	Upper body	Rehabilitation	Physical	Dysphagia	N = 3	YES	4 weeks; 5×/week; 12.5 min/session; 3 games/day
[57]	Upper body	Rehabilitation	Physical	Parkinson’s disease	N = 45 (15 IG [supervised]; 15 IG [non-supervised]; 15 CG)	YES	24 sessions; 3×/week; 75 min/session; 50–60 min playtime
[58]	Upper body	Rehabilitation	Physical	Stroke	N = 4	YES	12 weeks; max 6 games/day
[59]	Upper body	Rehabilitation	Physical	Stroke	N = 3	YES	6 weeks; 4×/week; 30 min/session
[60]	Upper body	Rehabilitation	Physical	Stroke	N = 8	YES	6 weeks; 18 sessions; 3×/week; 30 min/session
[61]	Upper body	Rehabilitation	Physical	Post-Traumatic hand injuries	N = 30 (15 IG; 15 CG)	YES	3 weeks; 15 sessions; 5×/week; 25 min/session
[62]	Upper body	Rehabilitation	Physical	Multiple sclerosis	N = 16 (10 IG; 6 CG)	YES	4 weeks; 12 sessions; 3×/week; 40 min/session
[63]	Upper body	Rehabilitation	Physical	Stroke	N = 12 (9 completed)	YES	10 sessions; 2–3×/week; 30 min/session
[64]	Upper body	Rehabilitation	Physical	Stroke	N = 15 (5 IG-MTD; 5 IG-MTD+EEG; 5 CG)	YES	4 weeks; 12 sessions; 3×/week; ~35 min/session
[65]	Upper body	Rehabilitation	Physical	Neurological disorders	N = 6	YES	1 week; 2 sessions; 40 min/session
[66]	Upper body	Rehabilitation	Physical	Chronic low back pain	N = 10	YES	6–9 sessions; 30–45 min/session; sessions 2–3 days apart
[67]	Lower body	Rehabilitation	Physical	Chronic kidney disease	N = 33	YES	12 weeks; 3×/week; 40 min/session
[68]	Full body	Rehabilitation	Physical	Parkinson’s disease	N = 10	YES	6–9 weeks; 18 sessions; progressive 15–40 min/session
[69]	Full body	Rehabilitation	Physical	Parkinson’s disease	N = 50 (25 IG; 25 CG)	YES	6–9 weeks; 18 sessions; 2–3×/week; progressive 15–45 min/session
[70]	Lower body	Rehabilitation	Physical	Parkinson’s disease	N = 7	YES	5 weeks; 20 sessions; 4×/week; 30 min/session
[71]	Upper body	Rehabilitation	Physical	Stroke	N = 4	YES	6 weeks; target ≥4 h/week; variable session duration; 2.5–3.5 h/week actual
[72]	Lower body	Training	Physical	Older adults	N = 7	NO	3 weeks; 3 sessions; 1×/week; 120 squats/session
[73]	Lower body	Rehabilitation	Physical	Gait/Balance impairment	N = 27 (14 IG; 13 CG)	NO	2 weeks; 3 sessions; 15–20 min/session
[74]	Full body	Rehabilitation	Cognitive	Autism Spectrum Disorder (ASD)	N = 6	YES	10 weeks; 10 sessions; 1×/week; 45 min/session
[75]	Lower body	Rehabilitation	Physical	Balance disorders	N = 4	YES	1 month; multiple sessions
[76]	Lower body	Rehabilitation	Physical; Cognitive	Gait/Balance impairment	N = 19	YES	1–5 sessions/participant; 25 min/session
[77]	Full body	Rehabilitation	Physical	Parkinson’s disease	N = 23 (11 IG; 12 CG)	YES	6–9 weeks; 18 sessions
[78]	Upper body	Rehabilitation	Cognitive	Stroke	N = 42 (21 IG; 21 CG)	YES	4–6 weeks; 18–25 sessions; 3–5×/week; 30 min/session
[79]	Full body	Rehabilitation	Physical	Chronic back pain	N = 22 (11 IG; 11 CG)	YES	4 weeks; 12 sessions; 30 min/session
[80]	Full body	Training	Physical	Hypertension	N = 22	YES	1 week; 2 sessions; 25 min/session
[81]	Lower body	Rehabilitation	Physical	Total knee replacement	N = 27 (16 IG; 11 CG)	YES	6 days; daily; 40 min/session
[82]	Lower body	Rehabilitation	Physical	Total knee replacement	N = 34 (18 IG; 16 CG)	YES	1 week; daily sessions; 40 min/session; rehab + intervention
[83]	Upper body	Rehabilitation	Physical	Postoperative C5 palsy	N = 10 (4 IG; 6 CG)	YES	24 weeks; scheduled sessions; 60 min/session; postoperative follow-up
[84]	Upper body	Rehabilitation	Physical	Stroke	N = 20	YES	4 weeks; 16 sessions; 4×/week; 45 min/session
[85]	Upper body	Rehabilitation	Physical	Stroke	N = 15	YES	3 weeks; 9 sessions; 3×/week; 60 min/session
[86]	Lower body	Rehabilitation	Physical	Degenerative lumbar spine disease (DLD)	N = 14	YES	6 weeks; multi-session; 30 min/session
[87]	Upper body	Training	Cognitive	Healthy adults	N = 29	YES	1 month; 20 sessions
[88]	Full body	Rehabilitation	Physical	Coronary artery disease	N = 11	YES	6 months; 3×/week; variable session duration
[89]	Full body	Rehabilitation	Physical	Coronary artery disease	N = 33 (11 IG1; 11 IG2; 11 CG)	YES	6 months; 3×/week; 60 min/session; home-based
[90]	Upper body	Rehabilitation	Physical	Humerus tuberosity	N = 31 (1 target; 30 healthy)	NO	4 weeks; 160 tasks total
[91]	Upper body	Rehabilitation	Physical	Stroke	N = 11	YES	6 weeks; ~27 sessions; 30 min/session
[92]	Lower body	Rehabilitation	Physical	Healthy adults	N = 30 (15 IG; 15 CG)	YES	1 week; 2 sessions; 5 min/session
[93]	Lower body	Rehabilitation	Physical	Stroke	N = 22 (11 IG; 11 CG); 11 (preliminary)	YES	2 weeks; 10 sessions; 5×/week; 30 min/session

Table 4. Evaluation and reported outcomes for usability, acceptability and UX, and training effectiveness.

Study	Usability (How)	Usability Outcome	Acceptability and UX (How)	Acceptability and UX Outcome	Training Effectiveness (How)	Training Effectiveness Outcome
[33]	Not assessed	Not reported	Not assessed	Not reported	WMFT–TIME; performance metrics: Duration; Normalized Speed; Movement Arrest Period Ratio	In-game performance measures reflected upper-extremity motor function
[34]	Not assessed	Not reported	Not assessed	Not reported	FMA-UE; WMFT-TIME; WMFT-FAS	Upper-extremity motor function improved following virtual ADL practice
[35]	SUS; Custom post-intervention qualitative questionnaires	Participants reported high perceived usability and ease of interaction with the adaptive serious game	Custom post-intervention qualitative questionnaires	Users found the system engaging and acceptable for home-based neck pain rehabilitation	Visual Analog Scale (VAS) for pain intensity; Neck Disability Index (NDI); Range of Motion (ROM); posture and performance metrics	Pain intensity, neck disability, and cervical range of motion improved following the intervention
[36]	Post-intervention user survey; qualitative feedback	Participants reported the system was easy to use and satisfactory for home-based therapy	Post-intervention user survey; qualitative feedback; adherence	Users expressed satisfaction and willingness to continue using the system	McGill Pain Questionnaire; Numerical Rating Scale (NRS); Patient-Specific Functional Scale (PSFS); in-game data	Temporary pain relief and functional improvements were observed in some participants
[37]	Not assessed	Not reported	Not assessed	Not reported	FMA; Brunnstrom Recovery Stages; Modified Ashworth Scale; Box and Block Test (BBT); Motricity Index; Active ROM	Upper-extremity motor function and active range of motion improved following Kinect-based VR training
[38]	SUS; custom questionnaires and feedback	Therapists and patients reported good usability of the immersive rehabilitation framework	custom questionnaires and feedback	The system was well accepted by both therapists and patients for rehabilitation use	Therapist evaluation of rehabilitation progress	Therapist evaluations indicated progressive improvement in patients’ rehabilitation performance
[39]	SUS; custom questionnaires and feedback	Therapists and patients reported good usability of the immersive VR system	custom questionnaires and feedback	The system was well accepted by both therapists and patients	LSTM-RNN; therapist evaluation	Therapist evaluations and LSTM-RNN outputs indicated improvements in motor performance
[40]	Not assessed formally	Not reported	post-session appreciation feedback; reuse opinions; therapists feedback	Participants and therapists reported high system appreciation and willingness to reuse it	Performance and game metrics	Performance time across rehabilitation exercises decreased over sessions, indicating improved task execution.
[41]	Not assessed formally	Not reported	Qualitative feedback	Participants reported approval and acceptance of the training software	Unified Parkinson’s Disease Rating Scale (UPDRS) III; 5-step-walking; arm abduction angle; angle of camptocormia; in-game metrics	Improvements were observed in UPDRS III scores and clinical mobility measures
[42]	Not assessed	Not reported	Qualitative discussion	Participants reported positive perceptions of the exergaming training	Quiet standing testing; Limits of Stability test; Functional Balance Test	Exergaming improved static and dynamic balance; conventional training improved static balance
[43]	Not assessed	Not reported	Structured post-treatment Likert-scale questionnaire; qualitative feedback	Participants reported high acceptability of the immersive VR intervention	MoCA; Cognitive Assessment of Minnesota; Nottingham Extended ADL; Hospital Anxiety and Depression Scale (HADS); Quality of Life (EuroQoL); length of hospital stay	Improvements were observed in cognitive function, activities of daily living, quality of life, and hospital stay, particularly in more severely impaired participants
[44]	Not assessed	Not reported	Not assessed	Not reported	Functional Reach Test (FRT); Unipedal Stance Test; Four Square Step Test; physiological measures	Improvements were observed in functional reach and unipedal stance performance
[45]	Not assessed	Not reported	Not assessed	Not reported	Average using time; bi-weekly longitudinal comparisons	Average task completion time decreased over the training period and stabilized after the third week
[46]	Not assessed	Not reported	GEQ; Physical Exercise Adherence Questionnaire	Positive	Senior Fitness Test	Functional capacity improved, with greater gains in participants who played with peers
[47]	QUEST; SUS; UEQ-short; RTLX; qualitative feedback	Patients and therapists reported satisfactory usability across standardized questionnaires	QUEST; pAR; UEQ-short (UX related items)	Participants reported good acceptance and presence of the AR-based rehabilitation system	Not assessed	Not reported
[48]	USEQ; NASA-TLX	Usability and workload were rated positively using standardized questionnaires	VAS (motivation; satisfaction); Pittsburgh Participation to rehabilitation Scale	Motivation, satisfaction, and participation in rehabilitation were rated positively	FMA; kinematic metrics (ROM; max speed)	Clinical motor outcomes and kinematic measures showed heterogeneous changes across participants
[49]	Not assessed	Not reported	engagement and emotion survey; exit interview	High engagement and positive emotional responses to the VR game	Performance and game logs	Performance metrics showed improvements over the intervention period
[50]	Not assessed	Not reported	Engagement Questionnaire; qualitative feedback	Positive engagement and satisfaction with the rehabilitation game	Not assessed	Not reported
[51]	Not assessed	Not reported	Not assessed	Not reported	Oswestry Disability Index (ODI); Short Form 6D—Quality Adjusted Life Years	Disability and health-related quality of life improved following the VR-based intervention
[52]	SUS; qualitative interviews	Usability was rated positively based on questionnaire scores and interview feedback	UTAUT questionnaire; qualitative interviews	Acceptance and intention to use the exergames were rated positively	Not assessed	Not reported
[53]	SUS	Usability was rated positively based on standardized questionnaires	semi-structured qualitative interviews; adherence	Participants reported positive experiences and acceptance of the cognitive training platform	MoCA; Rey Auditory Verbal Learning Test; Trail Making Test (TMT) A and B; Clock Drawing Test; Digit Span Test (DST); Functional Activities Questionnaire; 15-item Geriatric Depression Scale; Instrumental Activities of Daily Living	Improvements were observed across multiple cognitive, functional, and emotional outcome measures
[54]	Semi-structured qualitative interviews; qualitative usability feedback	Usability was rated positively based on qualitative interviews and usability feedback	UTAUT questionnaire; adherence; semi-structured interviews	Acceptance, adherence, and intention to use were rated positively by participants	BBS; 30-Second Sit to Stand Test; TUG; FRT; MoCA; TMT A and B; Letter Digit Substitution Test; DST	Improvements were observed in balance, mobility, strength, and cognitive performance measures
[55]	UEQ	Usability was rated positively based on the related questionnaire item scores	UEQ; custom acceptability questionnaire; adherence	Acceptability and UX were rated positively based on the used questionnaires and adherence	TUG; 10-Meter Walk Test (10MWT); Five Times Sit to Stand Test; Mini Balance Evaluation Systems Test; UPDRS III; Lindop Parkinson’s Physiotherapy Assessment Scale; Interactive Walkway	Improvements were observed in gait, balance, mobility, and walking-adaptability measures
[56]	Not assessed	Not reported	VAS; enjoyment questionnaire	Participants rated the VR blowgun game as enjoyable	Jaw-opening force; tongue pressure; lip-closing strength; forced expiratory volume	Improvements were observed in oral muscle strength and respiratory function measures
[57]	Not assessed	Not reported	Not assessed	Not reported	Hand Active Sensation Test; Wrist Position Sense Test; BBT; Nine-Hole Peg Test; Grip and pinch strength	Improvements were observed in upper-limb sensory function, dexterity, and strength measures
[58]	Custom engagement questionnaire; qualitative feedback	Engagement with the system was rated positively	Custom engagement questionnaire; qualitative feedback	Positive engagement and acceptance of the VR game	Not assessed	Not reported
[59]	Custom usability/UX questionnaire	Usability was rated positively	Custom usability/UX questionnaire	Positive feedback regarding comfort and enjoyment	Action Research Arm Test; kinematic and force parameters; t-tests	Improvements were observed in upper-limb motor function and kinematic performance measures
[60]	Not assessed	Not reported	Not assessed	Not reported	FMA; Motor Assessment Scale; active ROM; Force-Extension; kinematic parameters (speed; acceleration)	Improvements were observed in upper-limb motor function, range of motion, force, and kinematic measures
[61]	Not assessed	Not reported	Post-intervention questionnaire	Positive acceptance of the virtual rehabilitation system	ROM; Grip strength	Improvements were observed in hand range of motion and grip strength
[62]	Not assessed	Not reported	Questionnaire for UX (motivation; satisfaction)	UX, motivation and satisfaction with the serious games were rated positively	9 Hole Peg Test; BBT; EQ-5D visual analogue scale (EQ-VAS); Short Form 12 (SF-12)	Improvements were observed in arm function and health-related quality of life measures
[63]	Self-reported questionnaire	Usability was rated positively	Self-reported questionnaire	Participants reported positive acceptance and satisfaction with the VR intervention	Action Research Arm Test; BBT; Modified Barthel Index (MBI)	Improvements were observed in upper-limb function and activities of daily living measures
[64]	Not assessed	Not reported	Not assessed	Not reported	FMA-UE pre-post	Improvements were observed in upper-limb motor function
[65]	Therapist usability evaluation; Qualitative observations	Therapists reported positive usability based on evaluations and qualitative observations	GEQ; in-game; post-game)	Patients reported positive game experience and acceptance	Not assessed	Not reported
[66]	Not assessed	Not reported	post-intervention questionnaire and qualitative feedback; adherence	Participants reported positive acceptance, enjoyment, and adherence to the VR intervention	daily: pain intensity (NRS); Tampa Scale for Kinesiophobia; Pain Catastrophizing Scale; Pain Anxiety Symptoms Scale; pre-post: ODI; Functional capacity; Perceptual awareness; Motor imagery accuracy/speed	Reductions were observed in pain intensity and pain catastrophizing, with improvements in functional capacity
[67]	Not assessed	Not reported	Adherence to intervention	High adherence to the VR exercise program	4-m gait speed; Short Physical Performance Battery; TUG; One-Legged Stance Test; Sit-to-Stand tests; 6-Minute Walk Test (6MWT)	Improvements were observed in gait speed, balance, mobility, lower-limb strength, and walking capacity
[68]	Custom questionnaire (interest, competence, difficulty); questionnaire for experience	Usability was rated positively	Visual analogue scale (interest, satisfaction, discomfort)	Participants reported positive interest and satisfaction, with low discomfort	Freezing of Gait Questionnaire (FOG-Q); Activity Balance Confidence; Gait and Balance Scale Part B (GABS-B); UPDRS II/III; gait initiation kinematics	Improvements were observed in freezing of gait, balance confidence, gait and balance scales, UPDRS scores, and gait initiation kinematics
[69]	Likert scales questionnaire	Usability was rated positively	Likert scales questionnaire	Acceptability of the exergaming intervention was rated positively by participants	Stand-Walk-Sit Test (SWST); UPDRS I–IV; GABS-B; New FOG-Q; Activity-specific Balance Scale; MoCA; HADS; Parkinson’s Disease Questionnaire (PDQ-39)	Mixed effects were reported across gait, balance, motor, cognitive, emotional, and quality-of-life measures
[70]	Qualitative feedback	Usability was rated positively	Qualitative feedback	Positive acceptance and engagement with the rehabilitation game	10MWT; Completion Time	Performance improved across assessed mobility and task execution measures
[71]	Game Experience Questionnaire (IBM-derived); SUS; semi-structured interviews	Usability was rated positively based on questionnaire scores and interview feedback	Semi-structured interviews (feasibility); adherence	Positive feasibility, and adherence to the intervention	FMA-UE; Canadian Occupational Performance Measure; Activities-specific Balance Confidence Scale; Stroke-Specific Quality of Life Scale; Minimal Detectable Change	Mixed changes were observed across clinical outcome measures relative to minimal detectable change thresholds
[72]	Not assessed	Not reported	Informal qualitative feedback (comfort, engagement)	Positive feedback about comfort and engagement	sEMG patterns (RF; VM; GA; VL; BF); knee indicators (knee shakiness; knee distance; squat depth)	Improvements were observed in muscle activation patterns and squat execution performance
[73]	Post-session Likert-scale questionnaire	Usability was rated positively	Post-session Likert-scale questionnaire; SAM	Positive experience, satisfaction, and affective responses	Videos analysis; group comparison t-tests	Higher active muscle effort was observed in the affect-adaptive condition compared with the control condition
[74]	Not assessed	Not reported	Questionnaires for parents and caregivers; qualitative acceptance reporting	Improved real-life street-crossing behavior, with high agreement between respondents	Street-crossing metrics; navigation metrics; safety-skills questionnaire	Navigation metrics improved, while street-crossing errors and questionnaire scores showed no significant overall change
[75]	Not assessed	Not reported	Qualitative participant feedback (motivation; engagement)	Increased motivation and engagement during the gamified balance training	Four Stages Balance Test; Correct Position time; Mean Deviation	Improvements in balance performance measures, including increased correct-position time and reduced mean deviation
[76]	Informal qualitative feedback (No difficulties Reported)	Participants were able to use the HMD-VR exergame without reported difficulties during gait training	PRS; Qualitative observations; semi-structured interviews	PRS scores and qualitative feedback indicated that the virtual environments were perceived as restorative and engaging	Not assessed	Not reported
[77]	Not assessed	Not reported	Not assessed	Not reported	SWST; UPDRS III; GABS-B; FOG-Q; MoCA; gait kinematics	Improvements were observed in clinical gait and balance measures and gait kinematics following the exergaming intervention
[78]	Likert-scale treatment satisfaction (FEPVR items); therapist observations	Participants were able to use the immersive VR system during training	Qualitative user feedback; FEPVR items; PANAS	High acceptance and positive affect during VR training	Tower or London-German; TMT A and B; Wechsler Memory Scale–Revised; EQ-5D VAS; Alters-Konzentrations-Test	Significant improvements were observed in executive function measures, with no significant changes in other outcomes
[79]	UEQ	UEQ results indicated above average to excellent UX across pragmatic and hedonic dimensions	TUI	TUI results indicated high perceived immersion and acceptance of the VR-based therapy	Pain NRS; Chronic Pain Grade Questionnaire; Hannover Functional Ability Questionnaire for measuring back pain-related disability; Tampa Scale of Kinesiophobia; SF-12	Reductions in pain intensity and improvements in functional capacity were observed, while fear-avoidance beliefs and quality-of-life measures showed no significant changes
[80]	TUI; NASA-TLX	Results indicated good usability with no significant differences between the two training types	TUI	TUI results indicated high acceptance and intention to use for both exergames	Not assessed	Not reported
[81]	SUS	SUS scores indicated good overall system usability	ARCS-based questionnaire (Attention, Relevance, Confidence, Satisfaction)	Questionnaire results indicated high system acceptance and motivation	Continuous Passive Motion; knee-bending angles	Improvements were observed in knee-bending angles compared with conventional rehabilitation
[82]	SUS	SUS scores indicated good overall system usability	Not assessed	Not reported	American Knee Society Score; Self-Efficacy for Rehabilitation	Improvements were observed in knee function and rehabilitation self-efficacy compared with conventional rehabilitation
[83]	Not assessed	Not reported	Not assessed	Not reported	Surface EMG (MVIC, %MVIC, and fatigue index); Medical Research Council; NDI; EuroQoL-5 Dimension (EQ-5D); VAS; HADS; game metrics	Improvements were observed in muscle activation efficiency, pain, quality of life, and psychological outcomes, along with improved game-based performance metrics
[84]	Not assessed	Not reported	IMI; compliance/adherence; qualitative user preference	High motivation and compliance, particularly in the multiuser condition	FMA-UE; total arm displacement per session; session duration	Upper-extremity motor function improved, with greater arm displacement and longer sessions
[85]	Weekly Survey; Final Survey	Responses indicated positive perceived usability of the VR system	VERGE Survey; Weekly Survey; Final Survey	Results indicated high acceptance, engagement, and willingness to use the system for therapy	Upper-limb kinematics from Xsens 3D motion tracking (total arm displacement; reach distance; hand elevation)	Upper-limb kinematic measures indicated substantial arm movement and reaching during VR training sessions
[86]	Not assessed	Not reported	Not assessed	Not reported	10MWT; 6MWT; Five times sit-to-stand test; TUG; Center of pressure; Muscle force; pain intensity VAS	Improvements were observed in walking performance, balance measures, muscle force, and pain intensity following the VR training
[87]	USEQ	High user satisfaction with the VR maze application.	Not assessed	Not reported	Maze task performance (response accuracy %, completion time; autonomous vs. manual)	Performance showed high response accuracy and efficient task completion
[88]	Custom post-intervention questionnaire	Responses indicated positive perceived usability	Custom post-intervention questionnaire	Responses indicated high acceptance of the system	Not assessed	Not reported
[89]	Not assessed	Not reported	adherence and compliance	Rates indicated good acceptance of the home-based program	Sit-to-stand test (30 s); accelerometer; stress tests	Lower-limb functional strength improved, with mixed effects on activity and exercise tolerance
[90]	Not assessed	Not reported	Not assessed	Not reported	CMS assessment scale; motion indicators; game metrics	Improvements were observed in shoulder function and motion indicators
[91]	Not assessed	Not reported	Structured patient interview; Usage metrics	Data indicated good acceptance and sustained use of the home-based system	FMA-UE; WMFT; IMU-derived kinematic metrics	Mixed results, with improvements in FMA-UE but no significant changes in WMFT
[92]	Not assessed	Not reported	Not assessed	Not reported	Step-aside reaction time (SART-standing; SART-walking); number of collisions; EMG metrics; ground reaction force	Step-aside reaction time decreased and collision avoidance performance improved following VR training
[93]	IMI; semi-structured interviews	Responses indicated high motivation and positive UX	IMI; semi-structured interviews; dropout rate	Results and low dropout rates indicated good acceptance of the intervention	FMA-LE; MBI; BBS; Single-Leg Stance time	Lower-extremity motor function, balance, and functional independence improved, following the intervention

Table 5. Evaluation and reported outcomes for cybersickness, system accuracy, and metrics validity.

Study	Cybersickness (How)	Cybersickness Outcome	System Accuracy (How)	System Accuracy Outcome	Metrics Validity Assessment	Metrics Validity Assessment Outcome
[33]	Not assessed	Not reported	Not assessed	Not reported	Spearman correlations between VOTA metrics and WMFT–TIME	VOTA-derived metrics showed criterion validity against the WMFT
[34]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[35]	Not formally assessed	Not reported	Posture-matching accuracy evaluation	Posture matching accurately tracked and assessed patient movements	Not assessed	Not reported
[36]	Adverse event monitoring	No major adverse effects reported	Not assessed	Not reported	Not assessed	Not reported
[37]	Adverse event monitoring	No adverse effects reported	Not assessed	Not reported	Not assessed	Not reported
[38]	Not formally assessed; qualitative	No major adverse effects reported	System calibration tests	Acceptable alignment between real and virtual body movements	Gated Recurrent Unit Recurrent Neural Network (GRU-RNN) compared to therapist scores	GRU-RNN performance scores were consistent with therapist-assigned evaluations
[39]	Not assessed	Not reported	Comparison of system-detected and patient reports	System-detected events were consistent with patient-reported interactions	LSTM-RNN compared to therapist scores	LSTM-RNN performance estimates aligned with therapist evaluations
[40]	Custom VR sickness score	Mild (mostly older participants)	Not assessed	Not reported	Not assessed	Not reported
[41]	Not assessed	Not reported	Not assessed	Not reported	Narrative comparison between Kinect game metrics and clinical mobility improvements	In-game metrics showed consistency with clinical mobility measures
[42]	Not assessed formally (adverse events monitoring)	No adverse effects reported	Not assessed	Not reported	Not assessed	Not reported
[43]	Qualitative monitoring (adverse event reporting)	Well tolerated; 1 transient dizziness/fatigue reported for each group	Not assessed	Not reported	Not assessed	Not reported
[44]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[45]	Not assessed	Not reported	Repeatability; reproducibility	Consistent measurements were obtained across repeated tests	Not assessed	Not reported
[46]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[47]	Not assessed formally (adverse events monitoring)	No serious adverse effects reported	Not assessed	Not reported	Not assessed	Not reported
[48]	Not assessed	Not reported	Not assessed	Not reported	Narrative consistency between instrumental assessments and FMA outcomes	In-game kinematic metrics were consistent with changes observed in FMA scores
[49]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[50]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[51]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[52]	VRSQ—quantitative	No or mild symptoms	Not assessed	Not reported	Not assessed	Not reported
[53]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[54]	VRSQ	Low VR sickness symptoms; no severe adverse effects	Not assessed	Not reported	Not assessed	Not reported
[55]	Adverse events monitoring	Limited adverse events	Not assessed	Not reported	Not assessed	Not reported
[56]	Adverse events monitoring	No adverse events reported	Not assessed	Not reported	Not assessed	Not reported
[57]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[58]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[59]	Not assessed	Not reported	Not assessed	Not reported	Correlation analysis between test scores and system outcome metrics	System outcome metrics showed correlation with Action Research Arm Test scores
[60]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[61]	Not assessed	Not reported	IMU vs. clinical goniometer comparison; cross-sensor interference evaluation; abduction/adduction testing; repeatability testing	Sensor measurements showed close agreement with clinical goniometer measurements and consistent performance across tests	Not assessed	Not reported
[62]	Adverse events monitoring	No adverse events reported	Not assessed	Not reported	Not assessed	Not reported
[63]	Adverse events monitoring	No related adverse events reported	Not assessed	Not reported	Not assessed	Not reported
[64]	Not assessed	Not reported	Not assessed	Not reported	Correlation analysis between FMA-UE, hit rate, and TBR	Improved motor function correlated with increased hit rate and decreased TBR.
[65]	Not assessed	Not reported	Qualitative evaluation of Kinect tracking performance	Kinect tracking performance showed limitations in accuracy under certain conditions	Not assessed	Not reported
[66]	Self-reported adverse effects	No adverse events reported	Not assessed	Not reported	Not assessed	Not reported
[67]	Adverse events monitoring	No adverse events reported	Not assessed	Not reported	Not assessed	Not reported
[68]	Adverse events monitoring	No related adverse events reported	Not assessed	Not reported	Not assessed	Not reported
[69]	Adverse events monitoring	No cybersickness adverse events reported	Not assessed	Not reported	Not assessed	Not reported
[70]	Adverse events monitoring	No adverse events reported	Not assessed	Not reported	Movement time consistency with Fitts’ Law	Movement time metrics were consistent with Fitts’ Law
[71]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[72]	Adverse events monitoring	No adverse events reported	Not assessed	Not reported	Not assessed	Not reported
[73]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[74]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[75]	Not assessed	Not reported	Comparison with expert clinician observations; Accuracy; Recall; F1-Score	System posture-error detection showed acceptable accuracy when compared with expert clinician observations	Correlation/regression analysis between system-derived metrics (Correct Position vs. Mean Deviation); Ordinary Least Squares (OLS)	A significant negative relationship was found between correct-position time and mean deviation using correlation and OLS regression analysis
[76]	Qualitative assessment	No cybersickness adverse events reported	Not assessed	Not reported	Not assessed	Not reported
[77]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[78]	Adverse event monitoring; self-report	No adverse effects reported	Not assessed	Not reported	Not assessed	Not reported
[79]	Adverse event monitoring	No adverse effects reported	Not assessed	Not reported	Not assessed	Not reported
[80]	SSQ in previous publication	Oculomotor symptoms were reported, particularly during endurance training	Not assessed	Not reported	Not assessed	Not reported
[81]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[82]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[83]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[84]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[85]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[86]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[87]	Not assessed	Not reported	Not assessed	Not reported	Correlation between response accuracy and memory performance	Maze task accuracy was correlated with memory performance
[88]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[89]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[90]	Not assessed	Not reported	Not assessed	Not reported	Prediction of CMS using MLP from motion indicators	Motion indicators predicted CMS assessment scores
[91]	Adverse event monitoring; self-report	No adverse effects reported	Not assessed	Not reported	Correlation analysis between FMA-UE and IMU-derived assessment metrics	IMU-derived kinematic metrics were correlated with FMA-UE scores
[92]	Not assessed	Not reported	Not assessed	Not reported	Not assessed	Not reported
[93]	Adverse event monitoring; self-report	No cybersickness adverse events reported	Not assessed	Not reported	Not assessed	Not reported

3.5. Synthesis of Results

The following subsections synthesize results across studies, focusing on XR system characteristics, study characteristics, and evaluation domains.

3.5.1. XR Modalities, Immersion Configurations, Visualization Devices, and Tracking Systems

Across the reviewed literature, the majority of studies employed VR-based interventions, accounting for around 92% of the total studies (n = 56). A small number of studies, or about 3%, employed combinations of AR and VR (n = 2), which makes the percentage of studies that used VR over 95% (n = 58). The AR-based interventions accounted for approximately another 3% (n = 2). MR systems were reported by one study, accounting for less than 2% of the studies. Figure 2 illustrates this distribution.

Figure 2. Distribution of XR modalities across included studies.

In terms of immersion level, the majority of the implementations used either non-immersive or fully immersive configurations. Around 54% of the studies (n = 33) used non-immersive setups, while 33% of the studies (n = 20) employed fully immersive configurations. Semi-immersive setups were uncommon and reported in approximately 6.5% of the studies (n = 4). A similar number of studies (6.5%, n = 4) implemented hybrid immersive configurations combining fully immersive with non-immersive configurations. Figure 3 shows this distribution.

Figure 3. Distribution of immersion levels across included studies.

The most frequently used visualization devices were screen-based displays, which appeared in approximately 57% of studies (n = 35). HMDs were used in about 36% of studies (n = 22), with the majority employing VR HMDs (n = 20) and a small number of them using AR or MR headsets (n = 2). Few studies employed hybrid immersive configurations combining screen-based displays and VR HMDs in the same intervention. Figure 4 summarizes the appearances of different tracking approaches across the included studies.

Figure 4. Distribution of type of visualization devices.

Regarding the motion tracking hardware, more than 93% of the included studies (n = 57) used a form of optical tracking, with several studies reporting multiple tracking approaches. The most common tracking method was the RGB-D camera, used in approximately 61% of the included studies (n = 37). Optical outside-in tracking was reported in about 25% of studies (n = 15), while optical inside-out tracking appeared in around 15% of studies (n = 9). Inertial-based tracking was used in about 11% of studies (n = 7), including one study using inertial-based tracking for a 3 Degrees of Freedom (3DoF) VR headset. Other infrequently reported tracking approaches included mechanical or robotic systems, EMG, force-based sensing, infrared tracking, and RGB cameras. Figure 5 summarizes the appearances of different tracking approaches across the included studies.

Figure 5. Frequency of tracking technologies across the included studies.

At the hardware level, the reviewed literature was dominated by a small number of commercially available platforms. The markerless RGB-D camera tracking for most of the studies was implemented using Microsoft Kinect sensors. Among VR HMDs, earlier studies employed outside-in configurations, with most of them relying on HMDs from the HTC Vive family that use base stations, whereas more recent studies increasingly adopted HMDs using inside-out, which is standalone, with most of these headsets being Oculus (and Meta) based platforms.

Across the reviewed time period, temporal trends were observed for levels of immersion and tracking approaches. Earlier studies more frequently used non-immersive, screen-based XR systems, while more recent studies employed increasingly fully immersive configurations. Figure 6 shows the temporal trends of the included studies by level of immersiveness. In 2015, 4 of the 5 included studies relied on non-immersive configurations, while 1 employed a semi-immersive setup. Likewise, in 2016, all five identified studies relied exclusively on non-immersive systems. By contrast, in 2024, about half of the nine studies employed fully immersive configurations, while in 2025, the identified studies used either fully immersive systems (n = 3) or hybrid configurations combining non-immersive and fully immersive elements (n = 1). These changes in immersion were accompanied by changes in visualization devices. Earlier studies relied to a large extent on screen-based displays, while later studies increasingly employed HMDs.

Figure 6. Temporal trends of the included studies by level of immersiveness. (a) Relative distribution of immersiveness expressed as percentages by year. (b) Number of studies per year by immersiveness. Data are aggregated across all three XR modalities (VR, AR, and MR).

Temporal analysis also revealed changes in tracking configurations used in immersive XR interventions. Earlier HMDs exclusively relied on outside-in tracking approaches. Inside-out tracking approaches were increasingly reported in studies from 2022 onward, while outside-in configurations were becoming less frequent over time. Figure 7 shows the temporal changes in the distribution of outside-in and inside-out tracking technologies over the reviewed period.

Figure 7. Temporal changes in the distribution of outside-in and inside-out HMDs.

For the visualization hardware used, there was a shift from monitor-based configurations toward HMD-based systems in later studies. Included studies from 2015 and 2016 were all screen-based. In contrast, by 2024 and 2025, there were more HMD configurations, with 2025 having no exclusively screen-based setups. Similarly, for the tracking hardware, RGB-D camera–based tracking (e.g., Kinect) dominated the earlier included studies. However, in 2025, there were no Kinect-only systems. More specifically, the included studies after 2022 seem to increasingly rely on HMD systems and their tracking methods, with the included 2025 studies all utilizing HMDs either as standalone or combined with Kinect.

3.5.2. Study and Intervention Characteristics

About half of the included studies (30 of 61) focused on the upper body, followed by full-body (19 of 61) and lower-body interventions (12 of 61). Most studies were related to rehabilitation (53 of 61), and 8 targeted training. For the targeted functions, most studies focused on physical functions (52 of 61), while 6 explored cognitive functions and 3 targeted a combination of physical and cognitive (motor–cognitive) functions.

The included studies addressed a variety of conditions and target populations. Stroke was the most frequently represented condition, appearing in approximately one-third of the studies (20 of 61), and was followed by Parkinson’s disease. Figure 8 illustrates the distribution of target populations and conditions across the included studies. For the majority of studies (58 of 61), the target population described in the study was the same as the sample used.

Figure 8. Distribution of target populations and conditions of the included studies.

Among studies addressing stroke, the majority focused on upper-body interventions (17 of 20) while only a small number focused on full-body (2 of 20) and lower-body functions (1 of 20). In contrast, for Parkinson’s disease, most focused on full-body interventions (7 of 9), with just one study each addressing the lower and upper body.

All the included studies involved multiple intervention sessions, with half of them (30 of 61) reporting durations of up to 1 month. Durations of up to 3 months were reported in 21 studies, while 7 studies extended beyond 3 months. For three of the studies, the exact intervention duration was unclear. Figure 9 illustrates the distribution of intervention durations across the four categories. About one third of the studies (20 of 61) reported a medium sample size of 21–40 participants, while the majority of the studies involved fewer than 20 participants (34 of 61), with 18 studies reporting sample sizes of 10 participants or fewer. Only 7 of the 61 studies reported larger sample sizes with more than 40 participants. Figure 10 shows the distribution of sample sizes.

Figure 9. Distribution of the intervention durations.

Figure 10. Distribution of the sample sizes grouped into 4 categories.

3.5.3. User-Related Evaluation Domains and Outcomes

Usability was assessed in 29 studies. Twenty-one of them employed custom evaluation approaches, such as qualitative interviews and custom questionnaires, with thirteen of them relying exclusively on custom evaluation methods. Among the standardized methods of evaluation, the most frequently used was the SUS, which was employed in nine studies. All the studies that assessed usability reported positive outcomes. Figure 11 illustrates the distribution of the usability evaluation methods.

Figure 11. Methods used for usability evaluation.

Acceptability and UX were assessed in 45 studies. Among these, 36 of the studies used custom evaluation approaches, such as qualitative interviews and custom questionnaires, with 30 of them relying exclusively on custom evaluation methods. Twelve studies assessed acceptability and UX using adherence, dropout rates, or usage metrics, and three studies employed the UEQ. Twenty different non-custom questionnaires or evaluation approaches were reported. All 45 studies assessing acceptability and UX reported positive outcomes.

The effectiveness of the interventions was assessed in 53 studies. Among these, 47 reported positive outcomes, while 6 reported mixed results. A total of 44 studies relied on standardized assessment tools, either alone or in combination with custom methods, whereas 9 studies relied exclusively on custom evaluation approaches such as game-derived performance metrics.

Adverse effects and cybersickness were assessed in 25 studies. Only three studies reported the use of standardized or validated cybersickness assessment tools such as the SSQ and the VRSQ. Three studies reported qualitative or custom cybersickness assessments, and nineteen studies reported adverse event or adverse effect monitoring.

Among the 23 studies that employed fully immersive VR interventions, 3 used standardized or validated tools, and 10 relied on qualitative methods such as custom VR sickness tools or adverse effect monitoring. One of the two semi-immersive VR studies monitored adverse effects. Out of the 34 non-immersive VR and AR configurations, 7 studies reported adverse effects or adverse event monitoring. Due to heterogeneity in reporting methods and outcome measures, direct comparison of cybersickness severity across immersion levels was not feasible.

3.5.4. System Related Evaluation Domains and Outcomes

System accuracy was assessed in seven studies. Six of these studies reported positive outcomes, while one reported mixed results due to tracking inaccuracy. Most evaluations were completed through qualitative or indirect comparisons rather than standardized or quantitative metrics. Metrics validity was assessed in 12 studies, with all reporting positive outcomes. The validity of the metrics was most commonly assessed through correlation or regression analyses by comparing system-derived metrics with established clinical scales, ratings, or performance measures. Narrative descriptions and inferred consistency with known theoretical or clinical benchmarks were only occasionally used.

4. Discussion

4.1. Interpretation of Results on XR System Configurations

This scoping review aimed to identify relevant studies and describe their main features. As such, the findings do not imply comparative superiority or evaluation of effectiveness. The synthesis of the included studies indicates the dominance of VR-based systems and optical tracking technologies in XR-based rehabilitation and training studies. This observation is consistent with prior work describing VR as having more clearly defined fields of application and research areas, while AR is less represented and still emerging [94]. On the other hand, MR has been described as conceptually and historically derived from VR [95]. Additionally, the greater use of VR-based systems compared to AR is consistent with broader trends shown in large-scale analyses, where VR-based publications outnumber those using AR [94].

In the reviewed studies, most of the configurations were either non-immersive or fully immersive. Semi-immersive configurations using large-screen displays and hybrid interventions were less common. For hardware, the reviewed studies relied mostly on commercially available platforms, such as Microsoft Kinect for RGB-D tracking and HMDs from the HTC Vive, and Oculus (later Meta) families of headsets. Previous studies have shown that non-immersive rehabilitation systems usually rely on RGB-D sensors such as Microsoft Kinect [96], while mainstream VR HMDs are used for immersive healthcare applications [97].

4.2. Temporal Trends for XR Configurations and Tracking Approaches

Within the reviewed time period, a change from non-immersive to fully immersive configurations was observed. This shift is consistent with advances in AR/VR display architectures, particularly improvements in near-eye display designs for AR/VR headsets [98].

A similar change was also observed for the tracking approaches. Earlier fully immersive systems relied exclusively on outside-in tracking, while later systems increasingly adopted inside-out tracking. The change from only outside-in to inside-out HMDs aligns with previous technical evaluations on the advantages and disadvantages of the two different approaches, where inside-out tracking allows simpler system setup by eliminating the need for external tracking devices such as base stations, while the tracking accuracy remains sufficient for most VR applications, and the cost becomes reduced [99].

Beyond the changes in tracking architectures that favor inside-out HMDs, the decrease in hardware costs combined with increased system functionality has been a wider trend, even though cost reporting is inconsistent in XR healthcare studies [100]. The lower cost, together with the advancements in hardware, may have played a role in the increased adoption of fully immersive standalone HMDs that were observed in more recent studies. However, formal cost-effectiveness analyses or even discussions about cost were rarely reported among the included studies. This limits the ability to assess the economic value of different XR configurations in rehabilitation settings. While the improved affordability likely contributes to the adoption trends, accessibility-focused research indicates that cost and setup complexity, together with other limiting factors such as usability and infrastructure-related constraints, remain key barriers to the adoption and use of immersive AR and VR systems [101].

4.3. Interpretation of Study and Intervention Characteristics

Most of the included studies focused on rehabilitation rather than training and primarily targeted physical functions and outcomes instead of cognitive or a combination of them. The most frequently investigated condition in the included studies was stroke, followed by Parkinson’s disease. Post-stroke motor rehabilitation is one of the research areas where immersive XR is applied frequently [102]. The majority of the studies focused on upper-limb rehabilitation. Additionally, the focus of the target body area was often related to the condition, with stroke-related studies emphasizing upper limb rehabilitation, while Parkinson’s disease studies often employed full-body interventions.

The predominance of physically focused rehabilitation interventions could be related to the two most frequent conditions, as both stroke [103] and Parkinson’s disease [104] are associated with motor impairments. This may partly contribute to the higher number of motor rehabilitation applications. Additionally, motion-tracking technologies are designed to capture movement, which favors motor rehabilitation applications over cognitive interventions. This underrepresentation of cognitive-focused XR interventions compared to physical-based interventions may suggest future research opportunities to explore the potential benefits of immersive technologies in cognitive rehabilitation and training.

4.4. Interpretation of Evaluation Domains and Outcome Measures

The interventions presented in the included studies assessed outcomes extensively for usability, acceptability, and training effectiveness. However, usability and acceptability assessments relied heavily on custom and non-standardized tools. The use of standardized instruments such as the SUS can provide a better basis for comparisons across systems [105]. Overall, the use of custom and non-standardized tools for evaluation of usability and acceptability has also been noted in a previous study on telemedicine systems, which highlighted the recommendation for use of validated questionnaires such as SUS and TAM-2 [106].

Most of the included studies assessed the effectiveness of rehabilitation and training interventions. While the evaluation tools varied, depending on the focus of each study, the majority relied on standardized tools and only a few employed custom approaches for the evaluation. The reported outcomes were generally positive.

Cybersickness and related adverse effects present acceptability barriers to the adoption of immersive AR and VR systems [101]. Most of the included studies assessed adverse events or effects through monitoring or qualitative methods, rather than through standardized or validated tools for cybersickness. More than half of the fully immersive VR studies used some form of adverse effects or cybersickness monitoring, while only about one-fifth of the non-immersive VR and AR studies monitored adverse effects or events, which represent a broader category than cybersickness. Fully immersive VR studies were the only ones that made use of standardized or validated tools. However, most still relied on custom tools and adverse event or effect monitoring. The heterogeneity of custom evaluation approaches and the limited use of standardized cybersickness evaluation tools reduce the ability to directly compare the severity or incidence of cybersickness across immersion levels.

The assessment of cybersickness using standardized tools such as the SSQ, which was originally developed to quantify simulator sickness in flight simulators [31], or the VRSQ, which was specifically developed to assess VR sickness [32], is important, as cybersickness symptoms can manifest across multiple distinct dimensions, including nausea, disorientation, and oculomotor disturbance [107].

Only a small number of studies assessed system accuracy, and in most cases, this was completed through qualitative or indirect approaches. The relatively limited assessment of system accuracy was expected, as purely technical and design-centered studies without training or rehabilitation components were excluded. Metrics validity was assessed more frequently than system accuracy, often through correlation or regression analyses. Overall, studies that assessed system accuracy or metrics validity reported generally positive results.

4.5. Methodological Gaps and Research Considerations

This review highlighted several methodological gaps. A main concern is that many studies involved relatively small sample sizes and short intervention durations, which limits the ability for generalizations and long-term effects assessment. These study designs are more common for feasibility and pilot studies, while there is a lack of larger-scale studies.

Another methodological concern is the wide use of custom-developed assessment tools, particularly for usability and acceptability, which limits comparability across studies. The high diversity of different XR system configurations, intervention designs, and evaluation approaches further restricts the ability for broader comparisons. In addition, reporting certain critical study characteristics was occasionally incomplete. For example, some studies did not specify the duration of the intervention, while in others it was readily clear whether the intervention included commercial games.

Furthermore, most of the included studies were VR-based, with limited representation of AR and MR interventions. The reviewed literature was also concentrated on stroke and Parkinson’s disease, with approximately one third of the included studies focusing on stroke, while other conditions, such as autism spectrum disorder, chronic kidney disease, chronic neck pain, dysphagia, hypertension, mild cognitive impairment, multiple sclerosis, phantom limb pain, and palsy, were each represented by only one study. Most of the interventions focused on physical rehabilitation, whereas cognitive and motor–cognitive applications were less frequently addressed.

The characteristics observed in the reviewed studies point to several areas for future XR rehabilitation research. The underrepresentation of AR and MR approaches suggests research opportunities for use in areas that can benefit from their unique advantages, such as the integration with real-world environments to support task-oriented rehabilitation, real-time movement guidance, or home-based activity support in everyday contexts. In addition, the concentration of studies on pathologies such as stroke and Parkinson’s disease, and the predominance of physical-targeted interventions, suggests that a direction for future research could be the expansion of XR-based interventions to a broader range of clinical populations and to cognitive or motor–cognitive domains. Studies with larger sample sizes and longer intervention durations could help improve the generalizability of XR intervention findings. Similarly, the use of standardized and validated tools could improve reporting consistency and allow cross-study comparisons, particularly in areas such as the cybersickness assessment.

4.6. Limitations of the Scoping Review

There are a few limitations to this scoping review. The reviewed period was between 2015 and 2025, and extending this timeframe could potentially provide a better representation of longer-term changes and trends in the field.

Furthermore, only peer-reviewed journal articles were included. This may have led to the exclusion of innovative tracking-related studies published in other types of publications, such as conference proceedings. More specifically, due to the fast-moving pace of the XR research domains, the exclusion of conference proceedings may have limited the representation of emerging tracking innovations that are often first reported in conferences.

Although four major databases were searched, it is possible that some relevant studies were not captured. In addition, relevant studies might have been missed due to the search strategy.

The eligibility criteria applied to this review, like the focus on motion-tracking technologies, multi-session interventions, and the exclusion of conference proceedings and commercial off-the-shelf games, may have favored structured rehabilitation applications over training interventions, which were less frequently represented in the included studies.

Some system characteristics and tracking details had to be inferred when not explicitly reported. These inferences were based on contextual information, known device specifications, visual material within the studies, or the publication period. In addition, the classification of certain assessment tools was based primarily on how they were applied within the studies. These factors may introduce uncertainty.

This review did not include an assessment of risk of bias, as the focus was on mapping and characterizing the existing literature rather than comparing the effectiveness of the different approaches. Consequently, the effectiveness or superiority of specific XR modalities, tracking systems, or study characteristics was not evaluated. The findings of this scoping review should be interpreted as a descriptive mapping of the reported study characteristics, system configurations, and outcome domains, rather than a comparison aimed to determine superiority of certain research trends over others.

The categorization of outcomes (e.g., positive) reflects the reported results of the individual studies and does not establish any conclusions regarding the efficacy, effectiveness, or superiority of any methods or technologies.

Formal inter-rater agreement statistics, such as Cohen’s kappa, were not calculated. As a result, the precision with which reviewer consistency can be interpreted may be limited, despite the high percentage of agreement between the reviewers.

5. Conclusions

This scoping review mapped, identified, and synthesized 61 studies related to the use of XR systems for rehabilitation and training. The findings show that the research area has focused on VR-based systems. Variations were observed for the levels of immersion, visualization devices, and tracking methods. Most of the included studies targeted physical rehabilitation for stroke and Parkinson’s disease. Usability, acceptability and UX, and training effectiveness were assessed by most of the studies, and the reported results were generally positive. Study designs and evaluation approaches appeared to have significant heterogeneity. Future research could explore XR-based applications in rehabilitation and training across a broader range of clinical populations and cognitive domains. The findings also point to opportunities for further research on the implementation of AR and MR approaches, which were underrepresented, particularly in real-world and home-based contexts.