Stroke is a leading cause of long-term disability worldwide, with many survivors experiencing persistent gait and balance impairments that hinder independent mobility and daily activities (1). Chronic stroke patients often experience reduced walking speed, asymmetric gait patterns, and impaired balance control, which significantly increase the risk of falls and secondary health complications. These limitations affect physical activity and cardiovascular health, reducing social participation and lowering overall quality of life (2, 3). Thus, various rehabilitation approaches have been introduced to address these challenges. Treadmill-based gait training is widely used to improve walking speed, endurance, and lower limb coordination through repetitive practice (4, 5).

However, conventional treadmill training has inherent limitations, as it primarily focuses on linear, repetitive stepping movements without incorporating the adaptive strategies needed for real-world walking conditions. Effective ambulation requires the constant adjustments of environmental variables, such as obstacle negotiation, directional changes, and maintaining balance during multitasking. Traditional rehabilitation settings often lack the space and resources to simulate these real-world demands in a controlled and safe manner (6, 7).

Virtual reality (VR)-based treadmill gait training has been introduced as an advanced rehabilitation strategy to overcome these challenges. This training creates an interactive, immersive environment where patients can perform task-specific gait exercises with real-time visual feedback. By integrating augmented reality elements, such as obstacle avoidance, variable-speed walking, and balance challenges, VR-based treadmill training can improve motor learning and neuroplasticity through a progressive, adaptable training experience (8). Unlike traditional treadmill training, VR systems may allow patients to practice complex locomotor tasks in a controlled setting, helping them transfer these skills to daily walking activities in real environments (9).

Previous studies have demonstrated the effectiveness of VR-based treadmill training in improving gait function and balance in stroke patients. Heeren et al. (10) reported an open-label clinical trial with a one-group pretest-posttest study design in which VR-based training improved walking speed, balance, and confidence in individuals with chronic stroke. Van Ooijen et al. (11) reported another open-label clinical trial with a one-group pretest-posttest study design where C-mill training improved walking speed, obstacle avoidance and reduced attentional demands during walking. Timmermans et al. (12) reported a randomized controlled case–control study in which C-Mill VR+ training did not improve standard walking speed but improved context-specific walking speed compared to an overground FALLS program, and Yang et al. (13) found that augmented reality treadmill training significantly improved obstacle avoidance and balance, but 10-m walk test did not compared to regular treadmill programs in subacute stroke patients in a randomized, controlled case–control study.

However, the effect of VR-based treadmill training on poststroke patients in the chronic stage remains unclear because randomized control trials (RCTs) focusing on chronic stroke survivors who require targeted interventions to enhance gait adaptability and dynamic balance are lacking (14). Additionally, most prior studies only assessed overall gait and balance scores and obstacle avoidance without conducting detailed analyses of functional mobility, spatiotemporal parameters, and the symmetry of gait or postural stability (e.g., center of pressure velocity and limits of stability) (15).

Thus, this randomized controlled trial, using a prospective, open, blinded end-point (PROBE) design, aimed to comprehensively evaluate the effects of VR-based treadmill gait training using the C-Mill VR+ system on gait performance, functional mobility, and balance in individuals with chronic stroke and compare the clinical efficacy of VR-based treadmill training with conventional physical therapy.

This study aimed to comprehensively evaluate the effects of VTT on gait performance, functional mobility, and balance in chronic stroke survivors. The findings demonstrated significant improvements in gait performance, as reflected by increased walking speed (10MWT), enhanced walking endurance (6MWT), improved spatiotemporal gait parameters, and reduced gait asymmetry. Additionally, the VTT participants exhibited greater improvements in functional mobility (TUG, DGI, and POMA gait subscale), as well as enhanced balance (BBS and CoP sway velocity), compared to those receiving conventional therapy alone. These results support the potential of VTT as an effective rehabilitation intervention for improving functional gait and balance control in chronic stroke survivors.

This study results demonstrated significant improvements in gait performance following VTT, including increased walking speed (10MWT) and enhanced walking endurance (6MWT). It is consistent with previous RCTs that reported increased walking speed following VR-based treadmill training in stroke patients (13, 26, 27). Specifically, Wang et al. (26) demonstrated superior improvements in walking speed after VR treadmill training compared to conventional therapy. De Melo et al. (28) showed greater enhancements in walking endurance, assessed by the 6MWT in Parkinson’s disease patients who underwent VR treadmill training compared to those who received conventional therapy. These results highlight the beneficial effects of VR-based training on key gait performance parameters, suggesting improved functional walking capacity and endurance for daily activities and community ambulation.

This study also demonstrated significant improvements in gait symmetry in the VTT group, as evidenced by significant reductions in asymmetry indices for stance duration, swing duration, and single support time duration compared to the CT group. Previous studies have consistently reported improved gait symmetry following visual feedback treadmill training. Drużbicki et al. (29) reported that chronic stroke patients showed sustained improvements in stance and swing duration symmetry following visual biofeedback treadmill training. Similarly, Jo et al. (30) demonstrated significant reductions in step-length asymmetry in chronic stroke survivors after visual feedback training emphasizing anterior–posterior weight shifting. Although traditional gait training approaches may partially improve gait symmetry, they generally lack individualized, precise feedback mechanisms. Hesse et al. (31) reported limited and non-significant improvements in gait symmetry indices in stroke patients following conventional neurodevelopmental therapy. The limited effectiveness of conventional approaches may result from repetitive practice without personalized or precise visual and kinematic feedback. These findings support our hypothesis that real-time visual and kinematic feedback, as well as individualized and interactive gait tasks provided within the VR environment, actively promoted more symmetrical gait patterns by encouraging patients to utilize their affected limbs more effectively. Improvements in gait symmetry are clinically meaningful because symmetrical gait patterns are strongly associated with increased walking efficiency and reduced energy expenditure. Awad et al. (32) showed that decreasing step-length asymmetry significantly reduced the energy cost of walking in stroke patients, resulting in more efficient gait patterns. Wert et al. (33) further supported these findings by demonstrating that reduced energy expenditure during walking correlated with greater functional independence. Therefore, the significant improvements in gait symmetry observed in our study suggest that VTT may not only enhance the quality of gait but also contribute to long-term improvements in overall functional independence and mobility in chronic stroke patients.

Functional mobility, assessed through the TUG test, the DGI, and the POMA gait component, significantly improved after VTT in this study. Improvements in the TUG scores are particularly meaningful, as this measure closely reflects daily ambulatory activities involving transitional movements such as standing up, walking, turning, and sitting down. Previous studies consistently reported improvements in TUG performance following VR treadmill or gait training in stroke patients (10, 12, 26), Parkinson’s disease (34), and older adults (35). TUG performance is also closely associated with fall risk, with shorter completion times reflecting a reduced likelihood of future falls among older adults living in community settings (36). The significant TUG improvements observed in our study suggest that VTT effectively enhanced the participants’ abilities to perform complex, real-world functional movements safely and efficiently. Additionally, substantial improvements observed in the DGI scores and the gait component of the POMA in the VTT group indicate enhanced dynamic balance and gait adaptability. Timmermans et al. (12) previously reported significant improvements in context-specific obstacle negotiation using the C-Mill VR environment in stroke patients. Wang et al. (26) reported superior improvements in obstacle avoidance success rates on treadmill walking from VR-based training compared to conventional training. Similar improvements were observed in our study, highlighting the advantage of the VR training environment, which allowed patients to repeatedly practice realistic gait scenarios involving obstacle avoidance, variable walking speeds, and directional changes. Such training conditions likely facilitated improved gait adaptability, enabling the participants to transfer learned gait strategies effectively to real-life walking environments. These outcomes reinforce the importance of context-specific repetitive training in enhancing functional mobility, dynamic balance, and overall walking adaptability among chronic stroke survivors.

This study also demonstrated significant improvements in balance and postural stability following VTT. The participants in the VTT group showed greater improvements in functional balance measured by the BBS than those in the CT group. Additionally, significant reductions in sway velocity during challenging postural conditions, such as tandem stance, indicated enhanced dynamic balance control. These findings align closely with previous studies reporting similar improvements in balance and postural control in chronic stroke survivors who underwent VR treadmill training compared to conventional treadmill interventions (10, 37). Yang et al. (37) and Heeren et al. (10) reported greater improvements in dynamic balance and postural control in stroke patients assessed by the BBS following VR treadmill training compared to conventional treadmill training. Shin and Chung (38) demonstrated significant gains in balance functions when visual and auditory feedback were integrated into treadmill training. Collectively, these results suggest that the real-time visual and kinematic feedback provided by the VR environment facilitated more precise weight shifting and postural control strategies, contributing to improved balance performance observed in this study.

Notably, the significant reduction in CoP sway velocity during eyes-open conditions indicates that VTT specifically improved the patients’ abilities to effectively utilize visual feedback for postural stabilization. These findings are supported by Zhang et al. (39), who reported that visual feedback-based training improved weight distribution control and overall postural stability. Reduced CoP sway velocity observed in the tandem stance condition suggests enhanced core muscle stability and coordination ability from VR-based treadmill training. The task-specific challenges incorporated in the VTT, including obstacle negotiation and directional changes, likely promoted core muscle activation and improved overall trunk stability, both of which are critical for the successful execution of complex real-world gait tasks. Consistent with these findings, El-Nashar et al. (40) found that core-strengthening exercises significantly improved dynamic sitting balance and trunk stability in chronic stroke patients. Thus, the current findings highlight the benefit of VR treadmill training in enhancing trunk control and dynamic balance required during practical activities, such as turning and navigating obstacles. However, improvements in the LoS and the balance component of POMA under static balance conditions were not significant in this study. This limitation may be due to the VTT protocol, which primarily focused on dynamic gait tasks and visual information processing rather than static balance or sensory integration without visual cues. Hwang et al. (41) previously demonstrated that visual feedback training was effective in improving both static and dynamic balance in stroke patients by reducing CoP sway velocity and increasing LoS. Therefore, future VR-based training interventions should consider integrating proprioceptive and vestibular sensory inputs to further enhance comprehensive balance outcomes.

Although the post-intervention DGI approached the conventional fall-risk threshold, participants’ scores remained sufficiently below the scale maximum to rule out a true ceiling effect. The gain in DGI coincided with a reduction in CoP sway when visual input was available, while eyes-closed and LoS conditions showed no change. These findings suggest that the current VR protocol strengthened visually guided, feed-forward control of gait and posture rather than recalibrating proprioceptive or vestibular inputs. To promote broader multisensory integration and improve real-world transfer, future VR programs should incorporate tasks that challenge these additional sensory systems, such as reduced-vision walking, compliant or perturbing surfaces, and head-movement exercises.

Clinically, chronic stroke rehabilitation is often challenging due to the perceived plateau in recovery, whereby functional improvements tend to decrease over time (42). Nonetheless, this study provides strong evidence supporting that intensive, task-specific gait training through VR technology can promote neuroplasticity and facilitate motor relearning even in the chronic stage following stroke. Previous research by Teasell et al. (43) and Page et al. (44) indicated that targeted, intensive rehabilitation interventions could significantly enhance functional recovery in chronic stroke patients, regardless of time since stroke onset. Numerous other studies have consistently demonstrated the potential for meaningful improvements in long-term chronic stroke survivors (45, 46). The immersive and interactive characteristics of VR-based treadmill training enable the progressive, adaptive practice of various gait tasks, which may explain the significant gains in gait adaptability, symmetry, and postural control observed despite chronicity in this study. Furthermore, while prior studies largely focused on basic measures such as walking speed and endurance, the present study incorporated more comprehensive gait symmetry and postural stability assessments, thereby providing a broader understanding of gait recovery and functional restoration in chronic stroke patients.

In typical clinical settings, conventional gait rehabilitation often requires large spaces and carefully structured environments for the safe practice of real-world ambulation tasks, posing practical challenges for clinicians. VR-based treadmill training effectively overcomes these limitations by providing a controlled yet highly varied gait training environment within a limited physical space. Patients can repeatedly practice realistic walking scenarios, such as obstacle avoidance, speed variation, and directional changes, which are essential for transferring these acquired skills to real-world mobility. This is particularly beneficial for chronic stroke survivors who may encounter difficulty accessing community-based gait training due to environmental restrictions or fear of falling (47, 48). The study findings suggest that VR-based treadmill training can be a valuable rehabilitation tool for long-term stroke survivors with lower ambulatory function, providing a safe, engaging, and task-specific intervention.

From a physiological standpoint, four mutually reinforcing mechanisms may account for the superior gait gains produced by VR-based treadmill training. First, the coordinated delivery of optic flow, proprioceptive loading, and vestibular input strengthens multisensory integration within fronto-parietal–cerebellar networks, sharpening step-to-step error detection and correction (49). Second, the immersive, task-specific scenarios heighten attentional engagement and practice intensity; functional-MRI studies have consequently demonstrated up-regulation of the ipsilesional sensorimotor cortex and supplementary motor area after VR-enhanced gait practice (50). Third, the moderate-to-vigorous walking bouts intrinsic to VR treadmill sessions acutely elevate circulating brain-derived neurotrophic factor, a neurotrophin that supports synaptic remodeling and long-term motor recovery (51). Finally, repeated exposure to variable step lengths, obstacle negotiation, and speed shifts elicits peripheral neuromuscular adaptations—such as stronger dorsiflexor recruitment and reduced extensor co-contraction—that translate into more symmetrical and energy-efficient walking (52). Collectively, these interconnected pathways provide a coherent physiological rationale for the functional improvements observed in the present study.

However, several limitations warrant acknowledgment. First, this study assessed only the short-term effects of a 6-week intervention, and the long-term retention of these improvements remains unclear. Future research should incorporate follow-up assessments to determine whether gait and balance improvements are sustained over time and whether continued VR training is necessary for maintaining functional gains. Second, the study was conducted in a controlled clinical environment, where patients trained on a treadmill with visually simulated walking challenges. While this provided a safe and structured setting, it may not fully replicate the real-world ambulatory challenges that stroke survivors face. Future studies should investigate the effectiveness of VR-based training in overground walking environments, incorporating diverse terrains, cognitive dual-tasking, and unpredictable obstacles to improve ecological validity. Third, all participants in this study were outpatients, making it challenging to standardize the conventional therapy administered to both groups. Although efforts were made to ensure consistency in the intervention protocols, variations in therapist approaches and individualized therapeutic plans may have influenced the results. Future studies should consider a more controlled rehabilitation setting or a crossover study design to minimize potential variability in conventional therapy across participants. Furthermore, the total intervention time differed between groups (60 min in the VTT arm vs. 30 min in the CT arm). While this disparity reflects the add-on nature of our efficacy design, the unequal dose prevents complete separation of training-content effects from time-dose effects. A future time-matched randomized trial is therefore warranted to verify that the observed benefits are attributable to the virtual reality-based treadmill gait training compared to other training.

This study demonstrated the positive efficacy of VR-based treadmill gait training in improving gait performance, functional mobility, and balance in chronic stroke patients. Notably, significant improvements were observed even in long-term chronic stroke survivors with persistent gait impairments, suggesting the effectiveness of VR-based rehabilitation beyond the early recovery phase. The ability to simulate diverse virtual walking environments in a controlled setting highlights its potential to enhance functional mobility and reduce falling risks in real environments. Thus, VR-based treadmill training may serve as a valuable rehabilitation approach for chronic stroke patients; however, further research is needed to explore its long-term benefits and optimal implementation strategies.