Impact of Tai Ji Yunshou with Visual Feedback on Balance and Mobility in Stroke Patients

Introduction

Stroke is the second-leading cause of death and a major cause of adult disability worldwide [1]. Effective rehabilitation is crucial, particularly in the subacute phase of stroke phase (approximately 1 week to 3 months after onset) [2]. This early phase is a critical window for recovery, when neuroplasticity and spontaneous functional gains are maximized [3,4]. Optimizing therapy during the subacute phase of stroke can significantly influence long-term outcomes [5].

Among post-stroke impairments, balance dysfunction is especially prevalent-studies report that up to ~83% of stroke survivors experience balance deficits [6]. Such impairments contribute to a high fall risk, with approximately 25–40% of stroke patients having falls in the first months of recovery [7]. Falls not only cause injuries but also heighten fear of falling and limit mobility, impeding overall rehabilitation progress [8]. Postural control can be broadly categorized into static balance (maintaining stability in a stationary position) and dynamic balance (maintaining equilibrium during movements); both are often compromised after stroke and require targeted intervention [7].

Balance training is important for the functional rehabilitation of stroke patients in the subacute phase. Research indicates that both static and dynamic balance training are effective in improving balance and supporting functional movement recovery [9–11]. Compared to static balance training, dynamic balance training replicates more functional movements that are similar to those encountered in daily activities [12]. Tai Ji Yunshou is a core component of traditional Tai Ji and represents an accessible and effective dynamic balance training method. It features slow, coordinated, bilateral upper-limb movements synchronized with trunk rotation and continuous lateral weight shifting. As a low-impact and rhythmic exercise, Tai Ji Yunshou promotes sensorimotor integration, postural control, and attentional focus, making it well-suited for stroke rehabilitation. Studies have demonstrated that Tai Ji Yunshou improves balance and functional movement ability in patients during the subacute phase of stroke [13,14]. Specifically, 4–12 weeks of Tai Ji Yunshou training has been shown to enhance performance on the Berg Balance Scale (BBS), Single-leg Standing Test (SLST), Timed Up and Go Test (TUGT), and Barthel Index (BI) [15]. The slow and straightforward movements of Tai Ji Yunshou, which require only basic standing balance abilities, make it a practical and scalable method for improving balance in this population [13,14].

Survivors with the subacute phase of stroke usually present pronounced weight-bearing asymmetry during quiet standing, loading the paretic (hemiplegic) limb markedly less than the non-paretic side, as force-plate and plantar-pressure studies consistently show a center-of-pressure shift toward the unaffected limb within the first weeks to months after onset [16]. Uneven weight distribution can adversely impact their balance strategy [17]. Balance training with visual pressure feedback, which displays real-time weight distribution, has been shown to improve posture control, gait symmetry, and daily activity performance [18,19]. However, the lasting effectiveness of such concurrent feedback training remains a concern. Continuous concurrent visual feedback allows stroke survivors to correct their center-of-gravity distribution in real time; however, these improvements largely disappear once the display is removed, indicating poor retention and an over-reliance on visual cues [20,21]. Sensory-reweighting research further shows that prolonged dependence on real-time visual information prompts the central nervous system to down-weight proprioceptive, tactile, and vestibular inputs [20,22], so patients struggle to maintain postural gains after the feedback is withdrawn [23]. Consistent with motor learning theory, more durable balance improvements are achieved when rehabilitation protocols replace continuous feedback with lower-frequency, summary (terminal) feedback [24].

This study applied 2 forms of balance training – Tai Ji Yunshou and static standing – each combined with a visual feedback system that provided summative information about patients’ plantar pressure distribution during training. Unlike concurrent visual feedback, which displays real-time data, summative visual feedback refers to information presented after the completion of a training set, allowing patients to retrospectively observe and analyze their weight distribution patterns. This approach can reduce visual dependency and promote intrinsic sensorimotor learning [25]. Evaluation tools used included the BBS, TUGT, 10-Meter Walking Time (10MWT), BI, and lower-limb muscle strength testing. Although there is currently no consensus on the efficacy of combining static or dynamic balance training with summative visual feedback for patients in the subacute phase of stroke, the potential of this approach to enhance postural control and functional mobility warrants further investigation. Therefore, this study aimed to compare outcomes at 1 month on balance, walking, and muscle strength from Tai Ji Yunshou exercise with summative visual feedback, static standing with summative visual feedback, and conventional balance training in 80 patients in the subacute phase of a stroke.

Material and Methods

STUDY DESIGN, ETHICAL APPROVAL, AND REGISTRATION:

This prospectively designed, single-center, randomized controlled study was conducted at Renhe Hospital in Baoshan District, Shanghai, from January 2021 to April 2023. The study protocol was approved by the Institutional Review Board of Renhe Hospital (Approval No. 2020-33) on 25 September 2020, and the study was prospectively registered in the Chinese Clinical Trial Registry (Registration No. ChiCTR2000040083) on 20 November 2020, both prior to the enrolment of the first participant. All procedures were conducted in accordance with the ethical standards of the institutional and national research committees, the Declaration of Helsinki (as revised in 2013), and relevant regulatory requirements. Written informed consent was obtained from all participants prior to study inclusion, including consent for publication of any potentially identifiable data or images.

PARTICIPANTS:

The study included 84 patients in the subacute phase of stroke with hemiplegia, aged between 44 and 80 years. Inclusion criteria required patients to have a confirmed diagnosis of their first stroke by magnetic resonance imaging, obtained within a period of 2–12 weeks before initiating the study [26]. Patients were required to demonstrate unilateral functional and postural impairments related to the stroke, with Brunnstrom stage (BRS) ranging from IV to VI for the upper extremity, and III to VI for the lower extremity on the affected side [27]. Patients with vestibular dysfunctions, severe sensory deficits, or significant visual and auditory impairments that could affect their ability to follow verbal instructions or visual cues during balance training were excluded [15]. Additionally, patients with concurrent musculoskeletal or systemic diseases potentially affecting their participation in daily training activities were also excluded. Patients unable to walk at least 10 meters with or without walking aids, as well as those scoring below 32 on the BBS, were also excluded [28]. Furthermore, any individuals actively participating in self-directed or other balance training programs outside of this study were excluded.

SAMPLE SIZE CALCULATION:

An a-priori power analysis was performed with G*Power 3.1 (Heinrich-Heine University, Düsseldorf). Pilot data from 18 patients in the subacute phase of stroke (6 per group) showed mean±SD 4-week improvements in the BBS of 8.3±4.6, 5.1±4.4 and 3.0±4.0 for Groups 1–3, respectively. The corresponding one-way ANOVA effect-size index was Cohen’s f=0.368. Using a two-tailed α=0.05, desired power (1−β)=0.80, and 3 independent groups, the analysis yielded a minimum required total sample of N=75 (25 participants per group). If a 10% attrition allowance were applied, the recruitment target would rise to N=84 (≈28 per group).

RANDOMIZATION AND ALLOCATION CONCEALMENT:

A biostatistician not involved in treatment generated a computer-based permuted-block randomization list (block size=6; allocation ratio=1: 1: 1) with RANDOM.org. Each assignment was placed in a sequentially numbered, opaque, sealed envelope (SNOSE) by an independent research assistant, thereby concealing allocation until a participant completed baseline testing. Immediately after assessment, the enrolling investigator opened the next envelope to reveal group assignment, ensuring allocation concealment throughout recruitment. Eighty-five eligible participants were randomized: Group 1 (G1, n=28) – Tai Ji Yunshou with summative visual feedback; Group 2 (G2, n=28) – static standing with summative visual feedback; Group 3 (G3, n=28) – conventional balance training (no feedback). The enrolling investigator confirmed eligibility and obtained informed consent but took no part in treatment delivery or outcome assessment, maintaining separation between randomization and intervention procedures.

BLINDING:

Because of the nature of the interventions, therapists and participants could not be blinded. Outcome assessors and data analysts were blinded to group allocation. To maintain masking, assessment sessions were scheduled in a separate room and therapists were instructed not to disclose group identity.

THE BALANCE EVALUATION AND TRAINING SYSTEM:

The Balance Evaluation and Training System (NCC, China) includes independent left and right pedals designed to measure reactive forces (Figure 1). Each pedal is equipped with dual pressure sensors located at the forefoot and heel positions. Aligned with the zero-moment point, the midpoint between the pedals represents the true center of foot pressure. The system software, developed using Visual C++ 6.0, calculates the shifts in position and the trajectories of the center of pressure (COP). The COP derived from vertical forces, indicates the point of the ground reaction force vector. This reflects body sway and the forces that maintain the center of gravity (COG) within the base of support.

INTERVENTION PROTOCOL:

Participants in G1 and G2 received summative visual feedback balance training using a balance training system following their conventional rehabilitation sessions, while G3 received general balance training without visual feedback after the same conventional rehabilitation sessions. All groups trained for approximately 30 minutes per day, consisting of 10 sets of balance exercises lasting around 2 minutes each, interspersed with 1-minute rest intervals, 5 days per week over a 4-week period [10]. For G1 and G2, summative visual feedback was delivered via the Balance Evaluation and Training System. Although this system recorded plantar pressure data in real time, visual feedback was provided only after each 2-minute set of Tai Ji Yunshou (G1) or static standing balance exercise (G2). The feedback included a graphical display of the center of pressure (COP) trajectory and medio-lateral load distribution across the feet. Participants were instructed to review this information after each set to identify asymmetry and adjust their movements in subsequent sets. Importantly, no real-time correction cues were given during training, encouraging participants to engage in self-directed learning based on retrospective analysis.

CONVENTIONAL REHABILITATION (ALL GROUPS):

The rehabilitation program was uniformly administered to patients in all 3 groups by physical therapists, in accordance with established stroke rehabilitation guidelines [2]. The training included 1) A 15-minute warm-up exercises on a stationary bicycle. 2) Mobility and strength training for the upper and lower limbs, conducted in 3 sets of 15 repetitions. 3) Therapeutic exercises including active-assisted or passive movements for both lower and upper limbs, in 3 sets of 15 repetitions. 4) A 10-minute walking exercise session.

G1: TAI JI YUNSHOU WITH SUMMATIVE VISUAL FEEDBACK: Patients began in a natural standing position on the platform pedals, with feet at shoulder width and knees flexed at an angle between 5 and 30 degrees. Patients shifted their weight to the right leg, with the left hand ascending to face level, palm outward, and the right hand descending to hip level, palm down (Figure 2). This motion was mirrored during the weight transfers to the left leg, with the right hand elevating with palm outward, and left hand descending to hip level, palm down, to execute the Yunshou movement cycles. A 30-degree trunk rotation was targeted, guided by visual cues. Assistance was available for managing weight shifts. The training consisted of 8–10-second “Yunshou” cycles, totaling 10–12 sets in 2 minutes. During the training, the Balance Evaluation and Training System captured the path of the patient’s COG in real-time. Upon completing each set of Yunshou training, the system displayed a summary of the COG’s movement (Figure 3A). Patients reviewed the summary of the COG’s movement from their previous set. In the subsequent set, they were guided to adjust their posture and body weight shifts based on the outcomes from the previous set.

G2: STATIC STANDING WITH SUMMATIVE VISUAL FEEDBACK: The patients began in a natural standing position facing the monitor, with eyes open, arms by their sides, and legs straight, ensuring an upright posture for a duration of 120 seconds. During the training, Balance Evaluation and Training System captured the path of the patient’s COG in real time. Upon completion of each set of static training, the system displayed a summary of the COG’s movement (Figure 3B). Patients reviewed the summary of the COG’s movement from their previous set. In the subsequent set, they were guided to adjust their posture and body weight shifts based on the outcomes from the previous set.

G3: CONVENTIONAL BALANCE TRAINING: Participants in G3 received conventional balance training, which consisted of therapist-supervised postural control exercises commonly used in clinical stroke rehabilitation. These included static standing with eyes open, weight shifting in the anterior-posterior and medio-lateral directions, single-leg stance (as tolerated), heel-to-toe standing, and balance board activities. Each session lasted approximately 30 minutes, 5 times per week for 4 weeks, and no visual feedback (real-time or summative) was provided during or after the exercises [2].

PRIMARY OUTCOME MEASURE: Berg Balance Scale (BBS): The BBS consists of 14 items designed to assess static and dynamic balance abilities in individuals with neurological impairment. Each item is scored on a 5-point scale ranging from 0 (unable to perform) to 4 (normal performance), yielding a maximum total score of 56. Higher scores indicate better balance performance. The BBS has been demonstrated to have high reliability, validity, and responsiveness in individuals with stroke [29].

SECONDARY OUTCOME MEASURES: Timed Up and Go Test (TUGT): The TUGT evaluates functional mobility by measuring the time (in seconds) a participant takes to stand up from a seated position, walk 3 meters (approximately 10 feet), turn around, return, and sit down. Participants may use their usual walking aids during the test. Each participant performs the test 3 times with a 1-minute rest interval between trials. The final score is calculated as the average of 3 valid attempts. The TUGT has shown excellent test-retest reliability in stroke populations [30].

The 10-Meter Walking Time (10MWT): The 10MWT assesses walking speed over a short distance. Participants are instructed to walk 10 meters at a comfortable, self-selected pace. Timing is recorded for the 10 meters. The test is repeated 3 times with adequate rest between trials, and the average time is recorded. The 10MWT has demonstrated strong reliability and validity in post-stroke patients [31].

Barthel Index (BI): The BI measures functional independence in activities of daily living (ADL), including feeding, bathing, grooming, dressing, bowel and bladder management, toilet use, transfers, mobility, and stair climbing. The maximum score is 100, with higher scores reflecting greater independence. The BI is widely recognized as a reliable and valid outcome measure in stroke rehabilitation [32].

Muscle Strength Measurement: Lower-limb muscle strength was measured using the MicroFET2 handheld dynamometer, following standardized manual muscle testing protocols. The targeted muscles included the quadriceps (QUA), hamstrings (HM), tibialis anterior (TA), and gastrocnemius (GAS). Three trials were conducted per muscle group, and the average force value (in kilogram, kg) was recorded. The MicroFET2 has demonstrated good to excellent inter-rater and intra-rater reliability (ICC ≥0.70) in assessing lower-limb strength in stroke survivors [33].

STATISTICAL ANALYSIS:

All statistical analyses were performed using SPSS version 26.0 (IBM Corp., Armonk, NY, USA). The Shapiro-Wilk test was used to assess the normality of the data. Descriptive statistics were presented as means±standard error for continuous variables. One-way analysis of variance (ANOVA) was used to compare baseline characteristics and outcome measures among the 3 groups when data were normally distributed, followed by post hoc Bonferroni tests. For non-normally distributed data, the Kruskal-Wallis test was employed with Dunn-Bonferroni correction for pairwise comparisons. Within-group changes before and after intervention were analyzed using paired t-tests or Wilcoxon signed-rank tests, depending on the distribution. Statistical significance was set at a two-tailed P value <0.05 [34].

Result

PARTICIPANT FLOW AND BASELINE CHARACTERISTICS:

As shown in Figure 4, 84 patients were randomly assigned into 3 groups. Among them, 4 patients (4.8%) withdrew: 1 declined treatment, and 3 were discharged early. Thus, 80 patients (G1: n=28, G2: n=25, G3: n=27) completed the 4-week treatment protocol. No adverse events were reported. Baseline demographic and clinical characteristics – including age, sex, stroke duration, hemiplegic side, BRS, and scores for BBS, TUGT, 10MWT, and BI – are shown in Table 1, with no significant between-group differences.

BALANCE FUNCTION (BBS):

After 4 weeks, BBS scores improved significantly in all groups (P<0.001 for G1, G2, and G3). Between-group comparison showed a significant difference in BBS improvement (P=0.014, Table 2). Post hoc testing revealed that G1 showed significantly greater improvement than G3 (P=0.011, Table 3).

TUGT: After 4 weeks, significant within-group improvements were seen across all groups (P<0.001 for G1, G2, G3). However, no statistically significant between-group difference was detected (P=0.216, Table 2).

10MWT: After 4 weeks, all groups demonstrated significant within-group improvements (P<0.001). Between-group comparisons were significant (P=0.001), with G1 showing superior improvement compared to G3 (P=0.001, Table 3).

ACTIVITIES OF DAILY LIVING (BI):

After 4 weeks, all groups showed significant BI score improvements (P<0.001 for G1, G2, G3). Between-group differences were significant (P=0.003, Table 2). Post hoc analysis showed G1 had superior BI gains compared to G2 (P=0.041) and G3 (P=0.004), as shown in Table 3.

LOWER-LIMB MUSCLE STRENGTH:

After 4 weeks of intervention, all 3 groups showed significant within-group improvements in lower-limb muscle strength on both the hemiplegic and non-hemiplegic sides (all P values <0.001 unless otherwise stated; see Table 2). In particular, G1 exhibited the most pronounced gains in hemiplegic-side QUA, TA, and GAS strength compared to G2 and G3 (P<0.001 for QUA and TA vs both groups; P=0.005 vs G2 for GAS, P<0.001 vs G3 for GAS; Table 3). G2 also outperformed G3 in hemiplegic-side QUA strength (P<0.001).

On the non-hemiplegic side, between-group differences were observed primarily in the HAM and GAS. G3 showed greater gains than G1 in both HAM (P=0.005) and GAS (P=0.048), and better HAM strength than G2 (P=0.033). Interestingly, G2 demonstrated greater improvement in non-hemiplegic GAS strength compared to G1 (P=0.013).

Overall, the results suggest that G1’s combined Tai Ji Yunshou and summative visual feedback training led to more balanced and targeted strength gains on the paretic side, while G3 exhibited compensatory strengthening on the non-affected side, likely reflecting reliance on the unaffected limb.

Discussion

SUMMARY OF FINDINGS:

This randomized controlled study evaluated the effects of Tai Ji Yunshou training with summative visual feedback on patients in the subacute phase of stroke. After 4 weeks of intervention, all groups showed significant improvements in balance, gait, lower-limb muscle strength, and ADL. However, the group receiving Tai Ji Yunshou combined with summative pressure feedback (G1) demonstrated superior improvements across most outcomes, followed by the group receiving static standing with summative pressure feedback (G2), and lastly the conventional balance training group (G3). These findings suggest that integrating task-specific movement with low-frequency visual feedback may offer a synergistic advantage in promoting functional recovery after stroke.

BALANCE AND GAIT OUTCOMES:

Our results support the utility of Tai Ji Yunshou with summative visual feedback in balance rehabilitation. Notably, G1 achieved an 8.68-point increase in the Berg Balance Scale (BBS) and a 10.22-second decrease in the Timed Up and Go Test (TUGT), both exceeding the minimal detectable change thresholds [35]. These improvements are consistent with prior research, such as Kim et al’s study demonstrating the benefits of adding visual cues via laser-pointer feedback to trunk stabilization training [36]. Importantly, unlike concurrent feedback, our protocol utilized summative feedback delivered at the end of each set – encouraging reflective motor learning rather than dependence on real-time correction. Mechanistically, Tai Ji Yunshou offers 2 advantages documented in meta-analyses [14,37] and reflected in our data. First, its wide center-of-gravity excursions and 30° trunk rotations place continuous demands on postural control, activating trunk muscles and promoting dynamic stability. Second, the coordinated bilateral arm movements and lateral weight shifting engage upper- and lower-limb synergies, fostering proprioceptive awareness and functional integration. By pairing these whole-body challenges with terminal pressure feedback, patients were able to recalibrate asymmetrical loading strategies and improve postural control.

LOWER-LIMB MUSCLE STRENGTH AND SYMMETRY:

Muscle strength data further underscore the impact of Tai Ji Yunshou with summative visual feedback. G1 achieved the greatest gains in hemiplegic quadriceps, tibialis anterior, and gastrocnemius strength. In contrast, G3 (no feedback) showed greater gains on the non-hemiplegic side, suggesting compensatory overuse. These findings align with Hyun et al [18], who reported that real-time COP feedback during sit-to-stand training preferentially strengthened paretic lower-limb muscles. Our study extends these insights by demonstrating that delayed, set-summary feedback integrated into a bilateral motor task such as Tai Ji Yunshou can likewise promote symmetrical muscle recruitment. A 2023 meta-analysis by Zhang et al [13] confirmed that Tai Ji Yunshou leads to meaningful BBS improvements (~+4.9 points) and motor gains in stroke patients. Moreover, a 12-week study showed that body-weight-supported Yunshou improved paretic arm Fugl-Meyer scores more effectively than conventional therapy [38]. Together, these studies and our results support the view that bilateral, proprioceptive training – when paired with informative feedback – can drive neuromuscular rebalancing and durable motor recovery [39–42].

IMPLICATIONS FOR ACTIVITIES OF DAILY LIVING:

G1 resulted in the most substantial improvements in ADL, as assessed by the BI, with gains exceeding those in both the G2 and G3. The BI evaluates patients’ ability to perform essential functional tasks, including walking, transferring, stair climbing, and upper-limb activities such as eating, washing, and grooming [43]. The improvements observed in G1 likely stem from enhanced symmetry and coordinated activation of trunk and limb muscles, facilitated by the bilateral, dynamic nature of Tai Ji Yunshou movements and the integration of retrospective visual feedback. Tai Ji Yunshou requires synchronized circular arm motions combined with dynamic trunk and weight shifts, closely mirroring the motor demands of real-life ADL tasks [15]. This coordination-intensive approach may have contributed to the re-establishment of disrupted functional movement patterns by reinforcing proprioceptive control and multi-limb integration. These findings are consistent with previous reports demonstrating Tai Ji’s ability to improve motor performance in both upper and lower limbs among stroke patients [44,45]. The resemblance of Tai Ji Yunshou movements to real-world activities – such as reaching, lifting, or adjusting body posture – makes it particularly effective in promoting functional independence in the subacute phase of stroke. Moreover, as supported by prior studies [46,47], combining this traditional modality with summative visual feedback may further enhance neuromuscular adaptation and motor relearning. A study by Wiśniowska-Szurlej et al [48] showed that robotic-assisted gait training using the EksoNR exoskeleton significantly improved gait and functional status, underscoring a broader principle: both traditional, feedback-enhanced exercises and advanced technologies can yield meaningful gains in post-stroke recovery. Collectively, these findings highlight the importance of integrating innovative, task-specific, and feedback-driven strategies—such as Tai Ji Yunshou with summative feedback – into comprehensive stroke rehabilitation programs to optimize functional outcomes.

STUDY LIMITATIONS AND FUTURE DIRECTIONS:

This study has several important limitations. First, outcomes were assessed only at the 4-week post-intervention mark, limiting our ability to determine the long-term durability of the observed benefits. Second, the study design lacked 2 essential control conditions: a Tai Ji Yunshou-only group and a group receiving concurrent (real-time) feedback. The absence of these comparators prevents us from determining the independent effects of Tai Ji Yunshou and summative feedback, or whether their combined effect is additive or synergistic. Furthermore, we are unable to conclude whether summative feedback is more effective than real-time feedback in this population. From a methodological standpoint, the use of post-set (summative) feedback assumes a certain level of cognitive and perceptual processing, which may not be optimal for all subacute stroke patients, particularly those with visual or cognitive impairments. In addition, the Balance Evaluation and Training System used in this study captures center-of-pressure data but lacks detailed kinetic or kinematic information, limiting our understanding of underlying joint-level movement patterns.

Future studies should adopt a multi-center, factorial design to isolate the effects of Tai Ji Yunshou alone, summative feedback alone, concurrent feedback, and their combination. Extended follow-up (3–6 months) would help assess the retention of functional gains, and the inclusion of gait laboratory assessments or neuroimaging data could elucidate the neurophysiological mechanisms underpinning recovery. Such designs will be essential to establish the comparative and long-term value of integrating Tai Ji Yunshou with low-frequency visual feedback into post-stroke rehabilitation programs.

Conclusions

This randomized controlled study demonstrated that Tai Ji Yunshou exercise combined with summative visual feedback significantly improved balance, walking ability, lower-limb muscle strength, and activities of daily living in patients during the subacute phase of stroke, compared to both static standing with visual feedback and conventional balance training. The results suggest that incorporating dynamic, bilateral movement tasks with low-frequency visual feedback can enhance motor learning and functional recovery. Tai Ji Yunshou with summative visual feedback may therefore serve as an effective strategy in early post-stroke rehabilitation.