Between 2000 and 2021, a total of 1617 papers on VR rehabilitation were published. Emerging trends in VR rehabilitation research–related studies are outlined in Figure 2. From 2000 onward, VR rehabilitation research rapidly accelerated. In Figure 2, the blue line denotes the increasing trend in the annual number of studies from 2000 to 2021, whereas the red broken line represents the publication index, which also increased. In the initial 2000 to 2006 research period, publication numbers per year were relatively stable and were <25, suggesting an initial period of exploratory VR rehabilitation research. However, the rise in rehabilitation demands and the increased development of supporting technologies, which are due to the increased interest in applying VR to medicine and rehabilitation and the increasing, widespread use of VR technology, have resulted in a proliferation of publications since then. The number of publications in last decade accounted for more than 80% of the total publications found.

The analyzed publications included contributions from a total of 63 countries and 1921 institutes. The top 10 countries and institutes are outlined in Tables 1 and 2, respectively, whereas collaborations between countries and between institutes are shown in Figure 3 and Figure 4, respectively. The United States published the most studies (n=663), followed by Canada (n=143). The United States also had the top h-index (59), followed by Canada (29) and Australia (21). Rutgers State University New Brunswick had the most publications (n=52), followed by McGill University (n=45). Rutgers State University New Brunswick also had the highest h-index (27). On a global level, research institutes and associated research staff are collaborating and sharing experiences. However, although the United States has played a leading role, the international collaboration rates of the United States are low.

A total of 241 journals published articles related to VR rehabilitation; the top 10 are outlined in Table 3, and collaborations between cited journals are shown in Figure 5. Archives of Physical Medicine and Rehabilitation was the top journal (publications: n=847), followed by Physical Therapy and Rehabilitation (publications: n=608). Among these journals, Cochrane Database of Systematic Reviews had the highest impact factor (9.266), followed by Stroke (7.914).

Academic journals represent and reflect knowledge exchange in the research arena, where citing papers form a knowledge frontier and cited papers form a knowledge basis. A journal overlay dual map was assembled, as presented in Figure 6, wherein citing journals are shown on the left, cited journals are shown on the right, and citation relationships are represented by colored lines. These lines show that studies published in the molecular biology and genetics sector; the health, nursing, and medicine sector; the sports rehabilitation and sports sector; and the psychology, education, and social sector were usually cited in studies published in the medicine, medical, and clinical sector and the psychology, education, and health sector.

A citation is a vital bibliometric indicator, with frequently cited studies greatly influencing their research areas. We list the top 10 most highly cited publications in Table 4. Effectiveness of Virtual Reality Using Wii Gaming Technology in Stroke Rehabilitation by Saposnik et al [14], which was published on 2010 in Stroke, was the most highly cited article (cited 58 times); the article verified the effectiveness of rehabilitation for participants with stroke by using Wii (Nintendo Co, Ltd) gaming technology. Laver et al [15,16] published 2 articles with the same title—Virtual reality for stroke rehabilitation—on 2011 [15] and 2015 [16] in Cochrane Database of Systematic Reviews. These studies evaluated the rehabilitation impact of VR technology for patients with stroke in accordance with the status quo in different periods.

In network research, betweenness centrality reflects node importance in a network; thus, a higher betweenness centrality signifies an important study [23]. The betweenness centralities of the top 10 studies are shown in Table 4.

A cocited document–centered clustering investigation was proposed to identify subnodes and connecting nodes in VR rehabilitation. To assess the scientific relevance of publications, we established a network of cocited references (Figure 7). The cluster settings were as follows: the number of years per slice was set to 1, and the top 0.5% of articles were selected for analysis; a pruning algorithm was used. The modularity Q score was 0.8029 (>0.5), which showed that the network was well separated into loosely coupled clusters. The weighted mean silhouette score was 0.941 (>0.5) and suggested acceptable cluster homogeneity. Study index items were used as cluster markers. The largest cluster—cluster 0—was labeled as kinematics, cluster 1 was labeled as neurorehabilitation, cluster 2 was labeled as brain injury, cluster 3 was labeled as exergame, cluster 4 was labeled as aging, cluster 5 was labeled as motor rehabilitation, cluster 6 was labeled as mobility, cluster 7 was labeled as cerebral palsy, and cluster 8 was labeled as exercise intensity.

Keywords in similar publications were identified and processed. The top 20 keywords, which had a high link strength, are outlined in Table 5. As well as identifying thematic areas in this research field, our keyword examination of all articles (N=1617) identified 50 keywords with at least 16 occurrences (Figure 8).

We investigated hot spot shifts from a temporal perspective, using the top 2 keywords with the strongest citation burst—video games (2017-2021) and young adults (2018-2021; Figure 9).

This study analyzed 1617 SCIE papers that were published between 2000 and 2021 and were related to the research of VR rehabilitation. The United States published the highest number of studies (663/1617, 41.01%), and Canada published the second highest number of studies (143/1617, 8.84%). Most of the core research institutions in this field are from the United States and Israel. The US Food and Drug Administration's encouragement of digital therapy innovation is the main reason that the largest number of publications is in North American countries, led by the United States. Israel encourages entrepreneurship in the high-tech industry; as such, 4 of the top 10 universities in this field are in Israel. The most widely read journal was Journal of NeuroEngineering and Rehabilitation. These findings showed that Journal of NeuroEngineering and Rehabilitation significantly contributed to research in this area. Journal of NeuroEngineering and Rehabilitation focuses on the publication of research results in the fields of neuroscience, biomedical engineering, and rehabilitation, which is very much in line with the publication of papers in this field and has attracted the attention of relevant researchers. We also examined the top 10 cited publications; the top article by Saposnik et al [14] appeared in Stroke and was cited 58 times.

In previous studies, different VR-assisted rehabilitation systems were investigated in this field, with remarkable results. As indicated in Figure 7, after cocited reference clustering, key clustering nodes identified knowledge bases in this area, as follows: cluster 0 (kinematics), cluster 1 (neurorehabilitation), cluster 2 (brain injury), cluster 3 (exergame), cluster 4 (aging), cluster 5 (motor rehabilitation), cluster 6 (mobility), cluster 7 (cerebral palsy), and cluster 8 (exercise intensity). To this end, we described VR rehabilitation research knowledge bases by using different clusters, with time considerations.

For cluster 0, the kinematics cluster, scientists designed a VR rehabilitation system based on kinematics theory to measure, monitor, and predict outcomes; they achieved good results. For example, researchers developed various VR scenes for rehabilitation training for cervical vertebrae injuries; lower limb injuries of athletes; and arm, hand, and trunk movements in patients with stroke. They proposed kinematic evaluation and measurement methods based on the VR environment [24-26].

In cluster 1, the neurorehabilitation cluster, the research interest in VR technology for neurorehabilitation is increasing. Scientists have experimentally demonstrated the role of VR systems in neural rehabilitation and motor assistance from the perspectives of patients with impaired visuospatial perception after stroke, the augmented effect of intermittent theta burst stimulation on neurorehabilitation programs, and the use of motor imagination in brain-computer interfaces. Therefore, VR programs are considered safe and can be performed with standard neurorehabilitation protocols in patients with neurological conditions [27-29].

In cluster 2, the brain injury cluster, scientists demonstrated the experimental results of VR training systems for patients with brain injury, for those with acute onset central nervous system damage, and for vocational rehabilitation training from both the experimental perspective and the retrospective perspective. Importantly, most patients were responsive to VR training and showed improvements in gait function, balance control, vocational rehabilitation training, and other aspects [30-32].

In cluster 3, the exergame cluster, McMahon et al [33] investigated VR exergaming to augment physical activity in students (high school) with intellectual and developmental disabilities. Their findings indicated that when students used the VR exercise exergaming intervention, they increased the duration and intensity of their physical activity. Nambi et al [34], in their psychological and hormone analysis study, discovered that VR exergame training was effective for American soccer players with chronic low back pain when compared with conventional exercise training programs. In a multicenter controlled trial, Meyns et al [35] demonstrated that exergaming in children with spastic cerebral palsy improved balance after previous poor balance performances.

In cluster 4, the aging cluster, several investigations concentrated on upper limb rehabilitation intervention and overall skill improvement for older patients. Molina et al [36] analyzed the impacts of 13 VR sports games on physical function in older patients. Another group investigated the elbow and shoulder movements of older patients with chronic stroke, using the Predict Recovery Potential algorithm. The training improved motivation, but the benefits of physical function in aging were unknown [2,37].

In cluster 5, the motor rehabilitation cluster, scientists investigated various VR training devices. A team, for example, used VR-enhanced robot-assisted gait training to monitor the gait, motion, balance, fear of falling, and independence of 15 patients with chronic stroke. The researchers observed that cognitive performance and gait speed in tasks improved among study participants (P<.05) [38]. Maier at al [39], in their systematic investigation on randomized controlled trials, identified the benefits of specific VR systems for rehabilitating upper limb function and activity after stroke and reported that specific VR systems were highly beneficial for upper limb recovery when compared with conventional therapies.

In cluster 6, the mobility cluster, researchers designed remote VR and augmented reality rehabilitation training treadmills to record training data from patients with Parkinson disease or stroke. The experimental results showed that flexibility and balance improved and that VR rehabilitation methods could provide personalized rehabilitation strategies while also improving patient participation [40,41]. To identify the benefits of a VR-based interventional therapy, a previous scoping review reported that dynamic balance measures improved significantly following a therapeutic intervention; therefore, robust study designs that focus on intensity and dose responses from VR training could improve the efficacy of methods for treating mobility disorders [42].

In cluster 7, the cerebral palsy cluster, scientists studied gait, spatial perception exercises, functional mobility, exoskeletons, visuomotor construction, and other aspects of cerebral palsy in children by using data obtained through a VR rehabilitation training system. Bimanual performance and cognitive rehabilitation gains improved significantly during gait rehabilitation [43-45]. Furthermore, an augmented reality real-time feedback approach that incorporated infrared recognition technology was used to assess temporal and gait parameters during gait training in children with cerebral palsy. Importantly, cadence, velocity, functional ambulation, and bilateral step and stride length were augmented after the intervention [46].

For cluster 8, the exercise intensity cluster, scientists discovered that training factors (training intensity, prolonged activity, and rest were optimized) could potentially provide training optimization stimuli. de Vries et al [47] identified some elements of VR balance games that could potentially provide strength training stimulation. Baniña et al [48] investigated the relationship between upper limb motor recovery and exercise intensity in participants with subacute stroke. Importantly, they found that VR rehabilitation could be used to generate intensive exercise programs; however, clearer exercise progression guidelines should be published.

Keywords represent current research issues or topical concepts, while emerging trends and research frontiers are represented by burst keywords. We used the CiteSpace program to gather burst keywords and identified 2 scientific frontier areas with the strongest citation bursts—video games (2017-2021) and young adults (2018-2021).

In this study, we used a bibliometric analysis to objectively, comprehensively, and systematically analyze the VR rehabilitation research literature. We identified the knowledge bases, current topical hot spots, and oncoming trends in this area. In this research field, the areas with acceptable knowledge bases included kinematics, neurorehabilitation, brain injury, exergaming, aging, motor rehabilitation, mobility, cerebral palsy, and exercise intensity. We theorize that future emerging trends and frontiers will focus on video games and young adults. We also identified contemporary research topics and future trends in VR rehabilitation research and provided guidance for future research in this exciting area.

Although papers that were published at different times were gathered for this study, some were not comprehensive in nature and may have introduced publication bias, thereby affecting study outcomes.

Our study still has some limitations to be addressed. Although the WSCC database is reliable, other databases, such as non-English databases, should also be considered to ensure that all relevant papers are collected. Additionally, the rapid progress in the field of VR technology application limits the timeliness of this bibliometric study. Further, the VR-based rehabilitation activities we investigated included some gaming exercises, exergames, and video games that were not clearly defined. Finally, the results of the included studies only reflect current research trends in academia. In the future, VR technology is expected to be further integrated with clinical applications, which will enrich and improve the research in this subject field.