Comparative Effectiveness of Passive vs. Assistive Robotic Gait Training on Functional Recovery and Neuroplasticity Post-Stroke: A Randomized Controlled Trial

Comparative Effectiveness of Passive vs. Assistive Robotic Gait Training on Functional Recovery and Neuroplasticity Post-Stroke: A Randomized Controlled Trial | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Comparative Effectiveness of Passive vs. Assistive Robotic Gait Training on Functional Recovery and Neuroplasticity Post-Stroke: A Randomized Controlled Trial Yunjuan Xie, Meisi Song, Xiaodan Ma, Kaixuan Chen, Hui Xie, Kui Li, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6535268/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Robot-assisted gait training (RAGT) has gained recognition as a promising therapeutic approach, offering high-intensity and repetitive training. Despite its potential, the clinical effectiveness and ideal training protocols continue to be subjects of debate. This study seeks to compare the impacts of various RAGT modes on lower limb motor function recovery in stroke patients while exploring the corresponding neural mechanisms. Methods A double-blind, randomized controlled trial was conducted on patients aged 18 to 80 who had experienced their first unilateral stroke accompanied by walking impairments. Participants were randomly assigned to one of three groups: (1) assistive mode training combined with conventional therapies, (2) passive mode training combined with conventional therapies, or (3) a control group receiving only traditional rehabilitation. Outcomes were evaluated using the Fugl-Meyer Assessment for Lower Extremity (FMA-LE), Berg Balance Scale (BBS), Modified Barthel Index (MBI), the Functional Ambulatory Category (FAC), and functional near-infrared spectroscopy. Statistical analyses were performed using repeated measures ANOVA and non-parametric tests where appropriate, with statistical significance set at p < 0.05. Results Among the 48 patients recruited, significant time effects were observed across all groups in FMA-LE scores (p < 0.001). Notable improvements were detected in the conventional group (MD = 2.688, p = 0.005) and the passive group (MD = 3.667, p < 0.001), with the assistive mode also demonstrating a significant effect (MD = 1.789, p = 0.039). BBS scores improved across all groups; however, no significant differences were noted between the groups (p = 0.106). Similarly, MBI scores showed a significant time effect (p < 0.001), without notable group differences (p = 0.286). Crucially, the assistive mode training group exhibited significant differences in brain activation during tasks in specific regions compared to the control group, alongside notable interhemispheric connectivity differences. Conclusions All training modalities effectively enhanced motor function, balance, and daily living skills in stroke patients. Nevertheless, unique activation patterns and neural pathways suggest distinct underlying mechanisms for each training approach. Trial Registration: The study was registered with the China Clinical Trial Registration Center under the trial registration number ChiCTR2100054527. Robot-Assisted Gait Training Stroke Functional Near-Infrared Spectroscopy Motor Function Neuroplasticity Cortical Activation Functional Connectivity Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Stroke-induced motor deficits in the lower extremities profoundly affect patients’ ability to regain functional independence, participate in daily activities, and reintegrate into society. Consequently, gait rehabilitation constitutes a pivotal aspect of post-stroke recovery efforts [ 1 ]. Robot-assisted gait training (RAGT), utilizing innovations such as exoskeletons and end-effector systems, has emerged as a promising therapeutic approach, offering high-intensity and repetitive training. The efficacy of these robotic systems lies in their potential to enhance neuroplasticity, foster motor recovery, and improve mobility outcomes [ 2 , 3 ]. Evidence from systematic reviews and meta-analyses indicates that RAGT results in significant improvements in walking ability, speed, and balance, particularly when integrated with conventional therapies [ 4 , 5 ]. However, the comparative clinical effectiveness of RAGT versus traditional rehabilitation methods, along with the fine-tuning of training protocols, continues to be a subject of active debate within stroke rehabilitation research [ 4 , 6 ]. Over the past decade, significant progress has been made in developing robotic technologies and control strategies to optimize the interaction between patients and rehabilitation devices [ 7 , 8 ]. Assisted control strategies, which commonly involve trajectory tracking and compliant systems, are widely reported as prevalent methodologies in lower-limb exoskeletons [ 1 , 9 , 10 ]. The efficacy of these strategies varies across distinct training modalities: assistive paradigms promote neural plasticity through active volitional engagement, while passive modalities primarily leverage sensory afferent mechanisms to facilitate cortical reorganization. To reinforce the research rationale, it is essential to explicitly contrast the neuroplasticity mechanisms of active engagement in assistive mode with sensory dependence in passive mode. The assistive mode requires patients to volitionally initiate movement intentions, engaging the dorsolateral prefrontal cortex (DLPFC) to encode motor goals. These goals subsequently modulate primary motor cortex (M1) activity via thalamocortical pathways [ 11 ]. This process likely strengthens functional connectivity between the DLPFC and M1 through Hebbian plasticity, facilitating long-term cortical reorganization. In contrast, passive mode primarily relies on proprioceptive input, which activates the primary somatosensory cortex (S1) with minimal involvement of prefrontal cognitive processes [ 12 ]. Despite these distinct mechanisms, no research to date has directly compared the efficacy of assistive and passive mode training in terms of promoting neuroplasticity. Neuroimaging and neurophysiological research have concurrently offered valuable insights into the neuroplasticity mechanisms activated by RAGT. Studies utilizing functional near-infrared spectroscopy (fNIRS) and other imaging modalities have demonstrated increased cortical activation in stroke patients after undergoing robotic training, particularly in the ipsilesional motor cortex and associated brain regions [ 6 , 13 , 14 ]. The level of effort during robotic gait training has emerged as a pivotal factor, with evidence indicating that higher levels of effort during active walking activate broader cortical motor areas compared to fully assisted, passive walking in healthy individuals [ 15 ]. While these findings are promising, further research is needed to clarify the precise role of neuroplasticity in functional recovery and its relationship to various robotic systems and control strategies. Discrepancies in the dose-response relationship—specifically regarding training duration, frequency, and intensity—add complexity to the development of standardized treatment protocols [ 5 ]. Moreover, although some robotic systems surpass conventional physiotherapy in specific areas, such as improving gait symmetry, endurance, and balance, the results are not universally consistent [ 16 – 18 ]. The efficacy of RAGT also hinges on patient-specific factors, including the severity of motor impairment, ambulatory capacity, and the presence of comorbidities [ 18 , 19 ]. In conclusion, this article aims to examine the therapeutic efficacy of various RAGT modes on stroke patients, with a concentrated emphasis on the recovery of lower limb motor function and the neural mechanisms underpinning this improvement. Additionally, it seeks to evaluate the impact of different training modes on brain neuroplasticity, including potential variations in brain activation and functional connectivity. The central hypothesis suggests that assistive mode training—combining robotic assistance with the patient’s conscious effort to move—outperforms passive training in delivering therapeutic benefits. This approach is expected to promote greater brain activation, enhanced functional connectivity between motor and sensory regions, and more significant cortical reorganization. By refining treatment protocols based on these findings, there is promise for maximizing functional recovery, minimizing long-term disability, and enhancing the quality of life for individuals recovering from stroke. Methods Study Design and Participants This study utilized a single-center, double-blind, randomized controlled trial design. A random allocation sequence was generated using a random number table, and patients were divided into three groups: the assistive mode training group, the passive mode training group, and the conventional rehabilitation therapy group. The grouping details were securely placed inside opaque envelopes. During the trial, these envelopes were opened sequentially based on the patients' enrollment order, with smaller random numbers corresponding to earlier enrolled participants. Patients were then assigned to either the experimental or control group according to the instructions contained within the envelopes. A double-blind approach was implemented, ensuring that both the patients and the outcome evaluators were blinded, while the intervention personnel remained unblinded. With approval from the Ethics Committee of the Third Affiliated Hospital of Sun Yat-sen University, stroke patients with walking impairments hospitalized in the rehabilitation department of the same hospital and undergoing neurological rehabilitation training between September 2022 and May 2024 were recruited for the study. The trial was registered with the China Clinical Trial Registration Center under registration number ChiCTR2100054527. The study protocol adheres to the 2010 CONSORT guidelines for reporting parallel-group randomized trials [ 20 ]. The inclusion criteria specify patients aged 18 to 80 years who experienced their first unilateral stroke accompanied by walking dysfunction and provided informed consent, either personally or through their families. Exclusion criteria included individuals with non-stroke neurological conditions, such as craniocerebral trauma or neurocognitive disorders; those with cardiac pacemakers or intracranial metallic implants; patients with cranial defects or severe cognitive or communication impairments limiting emotional expression; individuals with a history of epilepsy or current pregnancy; and patients suffering from dysfunction of vital organs, including the heart, lungs, liver, or kidneys, or other severe systemic comorbidities incompatible with therapeutic interventions. All participants were fully briefed on the study's purpose and procedures and provided written informed consent before enrollment. Experimental Procedure Eligible participants were stratified and randomly assigned to one of three therapeutic cohorts: the assistive mode group, which received robot-assisted gait training in assistive mode combined with conventional rehabilitation therapies; the passive mode group, which underwent robot-assisted gait training in passive mode alongside conventional therapies; and the control group, which followed traditional rehabilitation protocols exclusively. All groups participated in interventions of equal duration, monitored by blinded assessors to ensure consistency. Patients in both the assistive and passive mode groups received treatment for 30 minutes per day, five days a week, for two weeks, totaling 10 sessions. The RAGT was conducted using the bilateral lower-limb exoskeleton system (AiWalker®, manufactured by Beijing AI-Robotics Technology Co., Ltd., China) (Fig. 1 ). This system integrates a closed-loop coordinated control mechanism to synchronize actuators at the bilateral hip and knee joints, ensuring alignment between the motor rotation centers and the anatomical axes of the corresponding joints. The exoskeleton is secured to the patient via adjustable straps positioned at the waist, thighs, calves, and feet. Key features include a fixed waist mechanism, powered lower-limb actuation, and support for overground walking. The waist support is equipped with adjustable movable linkages, enabling width customization to accommodate users with varying pelvic dimensions (measured as the distance between the left and right greater trochanters plus 2 cm). The system allows customization of hip and knee joint angles, offering maximal ranges of 33° flexion/23° extension for the hip and 53° flexion/full extension for the knee. Additionally, the gait cycle duration is adjustable, ranging between 2.45 and 5.25 seconds. The exoskeleton's anthropometric adaptability addresses multiple dimensions, including height, pelvic width, waist depth (defined as posterior protrusion from the midline), thigh length (measured from the greater trochanter to the knee joint gap plus 2 cm), and calf length (spanning the knee joint gap to the sole). For optimal use, patients must wear close-fitting clothing and flat-heeled shoes with heels shorter than 3 cm. Prior to usage, a pre-assessment of the lower-limb passive joint range of motion is mandatory. To ensure proper donning, the patient is seated with their feet positioned on the footplates, ensuring that the heels align with the posterior edges, while leaning securely against the backplate. Straps are systematically tightened at the waist, thighs, calves, and feet to ensure proper fixation. In the passive mode training group, the exoskeleton delivers complete kinematic assistance for ambulation without requiring voluntary engagement of the lower limbs. The telescoping rods for the thighs and calves enable swift and precise length adjustments to accommodate various anthropometric variations. Clinician supervision is advised to ensure both safety and the effectiveness of the training process. In assistive mode training, the patient actively generates force to initiate movement, which is augmented by a motor aligned with the knee joint that provides additional torque to aid the completion of the action. Assistive mode training evaluates the patient’s force intensity over three distinct intervals: the first 10 minutes, the middle 15 minutes, and the final 5 minutes. Training quantification utilizes Newton-based metrics for subjective effort assessment, supported by real-time sensor data acquisition. Clearly defined participation criteria require dynamic torque modulation to maintain a minimum of 60% volitional force contribution throughout the training sessions. All participants underwent standard inpatient rehabilitation during hospitalization, which included neurodevelopmental treatment based on the Bobath concept. This therapy was administered exclusively by certified physiotherapists accredited by the International Bobath Instructors Training Association. The core interventions were as follows: (1) Key point control to enhance dynamic stabilization (e.g., pelvis and scapular girdle stabilization); (2) Postural preparation training, such as trunk rotation with weight shifting while sitting and contralesional stepping during weight-bearing on the paretic limb; (3) Task-oriented exercises, including bridging drills to activate the gluteus maximus and transverse abdominis, sit-to-stand transfers facilitated by tactile cues (e.g., therapist’s hand placement on the knees to promote eccentric quadriceps control), and balance training on unstable surfaces coupled with visual distractions (e.g., ball-catching exercises); and (4) Gait re-education featuring body-weight support harnesses to refine segmented gait phases (e.g., heel strike, knee extension, and hip extension timing), paired with mirror feedback. Each session lasted 50 minutes and was conducted daily over a two-week period (10 sessions in total). The training parameters, such as surface stability and resistance levels, were customized to align with each patient’s functional capacity, following the methodology outlined by Louie et al. [ 21 ]. Clinical Assess Baseline data were collected prior to the initiation of the trial, capturing demographic variables (age, gender, height, weight, dominant hand) and clinical characteristics (lesion location, stroke type, time since stroke onset, stroke severity, and pre-intervention functional status). The primary outcome measure focused on lower limb motor recovery, assessed using the Fugl-Meyer Assessment for the Lower Extremity (FMA-LE). Secondary outcome measures included the Berg Balance Scale (BBS), Modified Barthel Index (MBI), Functional Ambulation Category (FAC), and corticospinal excitability parameters, such as resting motor threshold (RMT), motor-evoked potential (MEP) latency, and MEP amplitude. Blinded assessments were carried out at two time points: pre-intervention (T0) and post-2-week intervention (T1). These evaluations were performed by two independent physical therapists who had undergone a standardized assessor training program to ensure interrater reliability. Additionally, safety monitoring was conducted throughout the trial to document any adverse events. fNIRS Data Acquisition and Processing fNIRS data acquisition was performed using the Nirsmart system (Danyang Huichuang Medical Equipment Co., Ltd., China), which is equipped with 18 emitter probes transmitting dual wavelengths (740 nm and 850 nm) and 16 detector probes arranged with a standardized source-detector separation of 30 mm. Optode placement adhered to the 10–20 international system, targeting specific brain regions, including the bilateral M1, S1, DLPFC, prefrontal cortex (PFC), superior frontal cortex (SFC), premotor cortex (PMC), and occipital cortex (OC) (Fig. 2 ). During operation, ensure that the optodes are vertically and securely attached to the scalp, with optical gel applied for participants with long hair to minimize refraction loss. The experimental environment should be maintained at a stable temperature (20–25°C) and free from light interference, with participants using a head support frame to reduce motion artifacts. Continuous signal recording was conducted at a sampling frequency of 10 Hz. Before data collection, a 10-minute resting-state baseline calibration was performed, with real-time monitoring of the signal quality index. Channels with poor signal quality were adjusted or excluded as necessary. After the experiment, the optodes should be thoroughly cleaned, and the intensity of the light source should be periodically inspected. The raw optical intensity signals are converted into oxygenated hemoglobin (HbO) concentrations using a modified Beer-Lambert law [ 22 ]. Motion artifacts are addressed through the following steps: (1) artifact detection based on a moving standard deviation, identifying deviations exceeding 10% of the baseline; (2) cubic spline interpolation to correct transient artifacts lasting less than 2 seconds; (3) noise removal using principal component analysis (PCA), where the top three components accounting for over 95% of the variance are extracted; and (4) physiological noise reduction implemented with Butterworth bandpass filtering (0.01–0.5 Hz) and independent component analysis (ICA). After segmentation, the data undergoes baseline normalization using z-score standardization. Subsequently, the HbO signals are analyzed in the time-frequency domain through Morlet wavelet transformation within a 4–60-second window to isolate 0.01–0.08 Hz band power during task phases. Wavelet Amplitude (WA) represents a precise quantitative metric for assessing cortical activation intensity. The computational process unfolds as follows: Pre-processed time-domain signals obtained from fNIRS recordings undergo continuous wavelet transformation, generating time-frequency representations. This approach facilitates the temporal integration of frequency-specific components, producing time-resolved WA values. In essence, WA reflects the magnitude of task-induced hemoglobin oscillatory amplitudes within the cortical microvasculature, where higher WA values directly indicate enhanced neurovascular coupling responses in specific cerebral regions during task execution. Notably, robust WA measurements signify heightened functional specificity between cortical activation patterns and the cognitive-motor demands at hand. Functional connectivity has long served as a cornerstone metric for exploring inter-regional cerebral interactions, with its innovative methodological approach epitomized by the wavelet phase coherence (WPCO) algorithm [ 23 ]. Central to this framework is the decomposition of raw signals into time-frequency representations via continuous wavelet transform, which facilitates the extraction of instantaneous phase information across oscillatory frequency bands. The analytical pipeline unfolds across three distinct tiers: First, dynamic phase sequences are constructed from time-varying frequency components. Subsequently, the temporal coherence of inter-signal phase differences is quantified to yield WPCO coefficients—a parametric measure of cross-regional rhythmic synchronization, where higher values signify intensified neural coordination. In the final tier, an amplitude-adaptive Fourier transform technique is employed to mitigate amplitude-mediated phase distortion. This step effectively eliminates artifactual connectivity, isolating neurophysiologically valid functional coupling strengths that accurately reflect genuine neural activity dynamics. Statistical Analysis This study employed a three-arm randomized controlled trial with a repeated measures design, featuring two assessment points (pre- and post-intervention) for each participant across all groups. The sample size was determined using G*Power 3.1 software to analyze the group-by-time interaction effect, which constituted the primary hypothesis. Calculations were based on the assumption of a large effect size (Cohen’s f = 0.6), a two-tailed type I error rate ( α = 0.05), a type II error rate ( β = 0.20) corresponding to a statistical power of 0.8, a three-group structure, two repeated measurements, an assumed correlation coefficient of 0.5 among repeated measures, and a non-sphericity correction factor ( ε = 1.0) under the assumption of sphericity. The results indicated that a minimum of 33 participants (11 per group) would be required. To accommodate an anticipated attrition rate of 10%, the final sample size was increased to 39 participants, with 13 individuals assigned to each group. This trial followed the Per-Protocol principle, focusing exclusively on patients who strictly adhered to the study protocol. These patients met the study design criteria, received all prescribed treatments, and completed follow-up as outlined. The normality of the data was evaluated using the D'Agostino-Pearson test. Quantitative data with a normal distribution were reported as mean ± standard deviation ( x̄ ± s ), while non-normally distributed data were presented as the median ( P 25 , P 75 ). For baseline comparisons, independent t-tests were conducted for continuous variables, and χ² tests were used for categorical variables. Outcome analyses of normally distributed quantitative data included homogeneity of variance and sphericity assessments. When assumptions were satisfied, repeated-measures analysis of variance (ANOVA) was performed, with treatment groups (passive, assistive, and conventional) as between-subject factors and measurement times (T0, T1) as within-subject factors. Statistically significant results were further examined using least significant difference (LSD) post-hoc tests. In cases where sphericity assumptions were violated, the Greenhouse-Geisser correction was applied. For non-normally distributed or heteroscedastic data, comparisons between groups were conducted using the Kruskal-Wallis H test, while within-group comparisons utilized Wilcoxon signed-rank tests. Statistical significance was defined as two-tailed p < 0.05. Adverse events were analyzed to determine their relevance to the trial and evaluate overall safety. Data processing and analysis were performed using SPSS 22.0 software. Results Participant Flow and Characteristics A total of 48 patients, meeting the inclusion criteria, were recruited from the inpatient department of The Third Affiliated Hospital of Sun Yat-sen University between September 2022 and May 2024. Participants were randomly assigned in a 1:1:1 ratio to one of three groups: the passive mode training group, the assistive mode training group, or the conventional therapy group. Four patients in the passive mode training group withdrew for personal reasons after the initial assessment but prior to the start of treatment. Consequently, 44 patients completed the two-week intervention and underwent outcome evaluations (Fig. 3 ). Most participants had experienced a disease duration of 1 to 6 months. No significant differences were noted among the groups with respect to demographic variables (age, gender, height, weight, dominant hand) or clinical characteristics, including lesion location, stroke type, time since stroke onset, stroke severity, and pre-intervention functional status (Table 1 ). Additionally, baseline outcome measurements revealed no intergroup differences (Table 2 ), ensuring data comparability. No adverse events were reported throughout the study. Table 1 Baseline Characteristics Assistive mode training group (n = 16) Passive mode training group (n = 12) Conventional therapies group (n = 16) Age, mean (SD), year 58.44 (8. 15) 58.11 (8.25) 56.31 (12.66) Male patients, No. (%) 11(68.75%) 6(50%) 10(62.50%) Weight, median (IQR), kg 65(55.25–68.75) 65(46.50–70.70) 61(56.18–66.80) Height, mean (SD), cm 168.00 (7.69) 164.11 (8.64) 161.25(5.60) Body mass index , 21.94(3.13) 22.35(4.54) 23.88(3.73) Ischemic, No. (%) 10(62.50%) 7(58.33%) 9(56.25%) Affected side, No. (%) Left Right 8(50%) 8(50%) 5(41.67%) 7(58.33%) 10(62.50%) 6(37.50%) Time from stroke, median (IQR), days 96(41-139.5) 42(31–86) 95(51.25–166.5) SD, Standard Deviation; IQR, interquartile range. Table 2 Longitudinal Trajectories of Neuromotor Performance Metrics and Functional Recovery Indices in Robotic-Assisted vs. Conventional Therapy Assistive mode training group (n = 16) Passive mode training group (n = 12) Conventional therapies group (n = 16) Between-group difference ( p ) FMA-LE (score) T0 T1 19.02 (5.32) 20.81(5.26) 13.33 (6.44) 17.00(7.47) 20.00 (7.98) 22.69(6.26) 0.059 1.000 Within-group difference ( p /95%CI) 0.039(0.107 to 3.471) <0.001(1.991 to 5.342) 0.005(0.950 to 4.425) BBS(score) T0 T1 29.38 (10.54) 34.70(10.87) 19.22 (16.07) 25.78(17.61) 31.50 (14.99) 37.19(13.10) 0.099 0.129 Within-group difference ( p /95%CI) <0.001(2.995 to 7.645) 0.005(2.559 to 10.552) <0.001(3.050 to 8.325) MBI(score) T0 T1 51.69(22.69) 63.19(21.38) 47.89(22.70) 53.44(21.86) 61.75(23.07) 67.69(19.83) 0.284 0.273 Within-group difference ( p /95%CI) <0.001(5.624 to 17.376) 0.088(-1.036 to 12.147) 0.005(2.039 to 9.836) FAC(score), median (IQR) T0 T1 0.5(0 ~ 3) 2(0.25 ~ 3) 1(0 ~ 1.5) 1(0 ~ 2.5) 1.5(0 ~ 3) 2(0.25 ~ 3) 0.714 0.528 Within-group difference ( p /95%CI) 0.015 0.081 0.014 RMT (%) T0 T1 24.53 (9.17) 23.80(10.14) 27.44 (5.25) 25.11(7.80) 25.63 (10.09) 24.69(9.37) 0.741 0.938 Within-group difference ( p /95%CI) 0.350(-2.361 to 0.894) 0.180(-6.00 to 1.333) 0.649(-5.242 to 3.367) MEP latency (ms) T0 T1 34.19 (5.43) 35.39(5.24) 34.46 (3.09) 30.72(4.21) 32.92(4.91) 31.26(5.34) 0.674 0.035 Within-group difference ( p /95%CI) 0.241(-0.905 to 3.313) 0.126(-8.800 to 1.309) 0.346(-5.316 to 1.983) MEP amplitude (µV) T0 T1 239.70 (90.09) 219.19(77.95) 195.58 (75.02) 258.27(127.51) 187.62 (61.64) 214.21(100.88) 0.152 0.544 Within-group difference ( p /95%CI) 0.506(-84.974 to 43.954) 0.209(-43.136 to 168.507) 0.331(-29.804 to 82.979) FMA-LE, Fugl-Meyer Assessment for Lower Extremity; BBS, Berg Balance Scale; MBI, Modified Barthel Index; FAC, Functional Ambulatory Category; RMT, resting motor threshold; MEP, motor evoked potential. Clinical Outcomes and Assessments Data on lower extremity motor performance, balance, gait patterns, and functional status parameters in daily living are visually represented in Fig. 4 , with corresponding numerical analyses detailed comprehensively in Table 2 . Repeated-measures ANOVA results for FMA-LE revealed the following: Mauchly’s test identified a violation of sphericity, necessitating Greenhouse-Geisser correction for degrees of freedom. No statistically significant time × group interaction was observed (p = 0.331), indicating consistent improvement trends across all three groups over time. Regarding main effects, a significant time effect was identified (F(1, 41) = 30.761, p < 0.001, partial η² = 0.447), highlighting substantial improvement in patient scores across all groups. Post-treatment scores revealed significant increases in the conventional therapies group [mean difference (MD) = 2.688, p = 0.005)] and the passive mode training group (MD = 3.667, p < 0.001), while the assistive mode training group demonstrated a highly significant effect (MD = 1.789, p = 0.039). A marginally significant between-group main effect was also noted (F(2, 41) = 2.874, p = 0.069, partial η² = 0.131), which may be attributed to the limited sample size, warranting further validation with larger study cohorts. Importantly, the observed within-group difference of 6.1 points in FMA-LE surpasses the established minimal clinically important difference (MCID) threshold of 6 points for chronic stroke [ 24 , 25 ], underscoring the clinically meaningful advantage of robotic assistive mode training. The repeated-measures ANOVA results for BBS scores across the three groups yielded the following observations: At both T0 and T1, all groups showed improvement in mean BBS scores over time. Post-correction analysis revealed no significant time × group interaction ( p = 0.825), indicating parallel improvement trajectories among the groups. A statistically significant main effect of time was identified (F(1, 41) = 57.110, p < 0.001, partial η² = 0.600), with substantial post-intervention score increases observed across all groups: conventional therapy group (MD = 5.688, p < 0.001), passive mode training group (MD = 6.556, p = 0.005), and assistive mode training group (MD = 5.320, p < 0.001). Notably, the passive mode training group demonstrated the most pronounced improvement. No statistically significant between-group main effect was detected ( p = 0.106), suggesting comparable efficacy among the intervention strategies at the analyzed time points. Furthermore, the inter-group BBS improvement exceeded the chronic stroke MCID threshold (≥ 3.2 points), indicating measurable clinical effects across all groups. These findings support the implementation of each intervention as an evidence-based balance recovery strategy [ 26 ]. The time × group interaction for MBI showed no statistical significance ( p = 0.167), indicating that the improvement trajectories over time were comparable across the three groups. However, a significant main effect of time was observed (F(1,41) = 26.464, p < 0.001, partial η² = 0.411), demonstrating substantial clinical score improvements across all participants. Post-hoc analyses revealed statistically significant improvements in the conventional group (MD = 5.938, p = 0.005) and the assistive mode group (MD = 11.5, p < 0.0001), while the passive mode group exhibited no significant change (MD = 5.556, p = 0.088). Furthermore, no significant differences between the groups were identified ( p = 0.286). At T0, no statistically significant differences in FAC scores were observed among the three groups ( p = 0.714). Similarly, no notable between-group differences were detected in FAC scores at T1 ( p = 0.528). Within-group comparisons showed that changes in FAC scores within each group did not achieve statistical significance. However, a stratified analysis based on baseline FAC scores revealed statistically significant differences among the three groups within the FAC levels 0–1 category. Notably, a clinically meaningful distinction emerged between the assistive mode group and the passive mode group, with a mean difference of 5.49 ( p = 0.029; 95% CI: 0.61 to 10.37). Corticospinal Excitability Parameters The repeated-measures ANOVA for RMT indicated preserved sphericity ( p > 0.05), showing no statistically significant time × group interaction ( p = 0.782) or between-group differences ( p = 0.602). Modest RMT improvements were noted across all groups at the two-week follow-up; however, intergroup variations remained non-significant ( p = 0.169). In the MEP latency analysis, the findings revealed non-significant between-group differences ( p = 0.147) and no notable temporal effects ( p = 0.149). Regarding MEP amplitude, no substantial improvements were observed in any group. The analysis confirmed compliance with sphericity ( p > 0.05), with non-significant time × group interaction ( p = 0.236), negligible between-group differences ( p = 0.446), and undetectable temporal variations ( p = 0.239). Cortical Activation and Functional Connectivity After two weeks of varied training modes, distinct activation patterns in task-related brain regions emerged across the groups. A horizontal comparison between the conventional therapy group and the assistive mode training group revealed significant differences in task-rest activation values within the following brain regions during task execution: the contralesional SFC (t = 2.407, p = 0.022), the contralesional OC (t = 2.169, p = 0.038), the ipsilesional M1 (t = 2.383, p = 0.024), and the ipsilesional S1 (t = 2.354, p = 0.025). These differences remained statistically significant even after applying Bonferroni correction. Conversely, no statistically significant activation differences were observed between the conventional therapy group and the passive mode group or between the assistive mode group and the passive mode group (p > 0.05). Longitudinal analysis revealed that patients in the assistive mode training group demonstrated notable neuroplasticity in task-related activation patterns following two weeks of RAGT. Specifically, activation intensity under assistive mode conditions showed significant changes from baseline in the bilateral SFC, bilateral PMC, the contralesional S1, and the ipsilesional M1 (all p < 0.05). In contrast, the control group and passive intervention group displayed no significant pre- and post-training changes in activation values during identical task-state conditions. Prior to the intervention, cross-phase analysis of neural activation patterns revealed a dominant activation within the ipsilesional hemisphere during task conditions for all patients (Fig. 5 ). Post-intervention assessments in the conventional group showed reduced activation across bilateral cortical networks under task-elicited conditions. Neuroimaging analyses after treatment identified a predominant activation pattern localized to the ipsilesional cerebral hemisphere ( p < 0.05). In contrast, the assistive mode group displayed notable neuro-rebalancing effects, characterized by a significant reduction in contralesional hemisphere activation compared to pre-training, alongside an increased proportion of activation in the primary functional areas of the ipsilesional hemisphere. Meanwhile, the passive mode group did not achieve statistically significant differences in task-rest activation across either hemisphere following training (Fig. 6 ). The characteristics of cortical functional network reorganization in stroke patients under different training modes are summarized as follows: When comparing the conventional group to the assistive mode group, significant differences were identified in interhemispheric functional connectivity. These differences specifically involve the contralesional PMC and ipsilesional M1 ( p = 0.036); the ipsilesional PFC and contralesional PMC ( p = 0.025); as well as the ipsilesional PFC and contralesional S1 ( p = 0.048). In terms of intra-hemispheric functional networks, statistical differences were observed between the two groups within the contralesional hemisphere, particularly in the PFC-SFC ( p = 0.037), PFC-S1 ( p = 0.024), and PFC-OC ( p = 0.042) connectivities. Similarly, differences were observed within the ipsilesional hemisphere, involving the SFC-S1 ( p = 0.027) and PMC-S1 ( p = 0.026) connectivities. Following assistive mode training, patients demonstrated neuroplastic changes. Notably, intra-hemispheric functional connectivity in the contralesional hemisphere increased significantly compared to baseline. Newly formed functional couplings were observed between the S1 and SFC ( p < 0.05), S1 and PMC ( p < 0.05), as well as S1 and M1 ( p < 0.05). In the ipsilesional hemisphere, statistically significant functional connectivity was noted between the PFC and DLPFC, SFC and PMC, SFC and S1, S1 and M1, and S1 and OC (all p < 0.05). Lateral comparison analysis between the conventional group and the passive mode group highlighted inter-group differences in interhemispheric functional connectivity, specifically between the contralesional PFC and ipsilesional DLPFC ( p = 0.045). After two weeks of passive training, patients in the passive mode group exhibited a distinct pattern of reorganization. Changes in interhemispheric connectivity were focused between the contralesional SFC and ipsilesional OC, as well as the contralesional S1 and ipsilesional M1. Intra-hemispheric connectivity changes were observed between the contralesional DLPFC and SFC, contralesional SFC and S1, ipsilesional SFC and PMC, and ipsilesional S1 and M1 (Fig. 7 ). Associations Between Motor Improvements and Neurophysiological Metrics The analysis of neurofunctional score–brain activation correlations revealed distinct patterns across groups. In the conventional group, higher MBI scores showed strong positive associations with increased activation in the contralesional S1 ( r = 0.785, p < 0.001), ipsilesional SFC ( r = 0.823, p < 0.001), and ipsilesional M1 ( r = 0.659, p = 0.006), reflecting large and moderate effect sizes, respectively. In contrast, the assistive mode group demonstrated significant MBI improvements accompanied by heightened neural synchronization in the ipsilesional PFC ( r = 0.580, p = 0.019), highlighting the role of prefrontal regulatory mechanisms in motor recovery. Meanwhile, the passive mode group exhibited a positive correlation between functional gains and contralesional OC activation ( r = 0.747, p = 0.021), pointing to the involvement of visual processing networks in compensatory strategies. Discussion This study systematically examines the differences in the roles and neural mechanisms of passive mode, assistive mode, and conventional rehabilitation training in the motor function recovery of stroke patients. It utilizes an integrative approach involving behavioral assessments (FMA-LE, BBS, MBI, FAC), neuroelectrophysiology (MEP), and functional imaging (WA, WPCO). The findings indicate that all training groups achieved sustained functional improvements over time, consistent with the "time-sensitive window" phenomenon of post-stroke neuroplasticity [ 27 , 28 ]. Significantly, variations in the efficacy of functional improvements were observed across the different training modes. While the passive mode group exhibited the greatest improvement in balance function, as reflected in BBS scores (MD = 6.556), the intergroup difference was not statistically significant ( p > 0.05). This suggests that balance function recovery may depend more on consistent training intensity rather than the specific training mode employed [ 29 – 31 ]. The observed effect is likely attributable to sustained proprioceptive input in the passive mode, which may enhance postural control abilities through the activation of the cerebellum-vestibular pathway [ 32 ]. Conversely, the assistive mode demonstrated distinct advantages in both balance function (BBS) and activities of daily living (MBI). Notably, MBI scores showed a highly significant increase (MD = 11.5, p < 0.0001), underscoring the critical importance of active participation in maximizing functional gains. This improvement is achieved through the reinforcement of closed-loop regulation within the motor planning and execution system [ 33 ]. For instance, prior studies have indicated that active training can induce motor cortex reorganization, providing an optimal neural environment for motor rehabilitation [ 34 , 35 ]. The FAC-stratified analysis revealed differences in treatment responses; however, these findings should be interpreted cautiously given the limited statistical power. The post hoc nature of the analysis, coupled with small sample sizes within each stratum, increases the risk of Type II errors, positioning these results as hypothesis-generating rather than definitive. The stratification of treatment efficacy by baseline injury severity offers deeper insights into the mechanisms of action of various training modalities. In the subgroup with severe baseline motor dysfunction (FAC grades 0–1), the assistive mode group demonstrated significantly greater FMA improvement compared to the passive mode group, offering valuable guidance for clinical decision-making. Neuroimaging evidence indicates that patients with severe dysfunction often struggle to voluntarily initiate effective movements due to disrupted motor network connectivity on the affected side [ 36 , 37 ]. Assistive training mitigates this challenge by reducing task difficulty through real-time robotic assistance, enabling patients to attempt active movements within a manageable range. This fosters a positive cycle of “task difficulty adaptation–affected-side network activation–synaptic efficacy enhancement,” ultimately helping to re-establish neural pathways between motor intention and execution [ 38 , 39 ]. This proposed mechanism is supported by near-infrared imaging, which shows enhanced functional connectivity between the PFC and M1 on the affected side in the assistive mode group [ 40 , 41 ]. Additionally, prior research suggests that compensatory activation of the unaffected side may inhibit the remodeling potential of the affected side [ 42 , 43 ]. Conversely, passive mode training proves less effective in the subgroup with severe dysfunction, as its reliance on external drive hinders cortical active reorganization [ 44 ]. Nonetheless, passive mode training retains value; through continuous mechanical assistance, it generates rhythmic joint movements that deliver substantial proprioceptive and tactile input to the central nervous system [ 45 ]. These sensory signals, transmitted via the thalamocortical pathways to premotor and sensory integration areas, play a critical role in refining motor planning and enhancing muscle coordination [ 46 ]. Even in the absence of active participation, patients in the passive mode group in this study improved balance control through visual feedback mechanisms. This effect may be attributed to compensatory activation of the occipital cortex on the unaffected side. As Krakauer (2006) noted, the visual system plays a pivotal role in motor control compensation, particularly when motor execution is impaired. Visual information can help reorganize motor function by providing spatial positioning cues [ 33 ]. Overall, these findings highlight the necessity for clinical rehabilitation practices to consider the distinct impacts of different training modalities on the sensorimotor integration system [ 47 ]. From the perspective of neural remodeling mechanisms, fNIRS demonstrates that the brain’s network reconstruction, induced by different training modes, displays distinct paradigm-dependent characteristics. Patients in the assistive mode group exhibited a pronounced affected-side dominant activation pattern following training, marked by significant activation ( p < 0.05) in the bilateral SFC, PMC, and affected-side M1 during task states [ 48 ]. Moreover, their improvement in MBI showed a significantly positive correlation with increased neural synchrony in the ipsilesional PFC. This observation aligns closely with the task-specific neural remodeling theory [ 49 ], suggesting that the assistive mode enhances functional coupling between prefrontal regulatory regions and motor execution areas, thereby facilitating the efficient conversion of motor intentions into actual movements [ 50 ]. In contrast, the passive mode group demonstrated a unique unaffected-side compensatory mechanism. In this group, the activation intensity of the occipital cortex correlated significantly with MBI improvement, likely driven by the continuous input of visual cues, such as real-time feedback on robotic movement trajectories during passive mode training. This process appears to promote adaptive reorganization within the visual-spatial processing network. Meanwhile, patients in the conventional group exhibited unaffected-side dominant activation patterns consistent with prior studies. However, their MBI improvement still showed a significant correlation with activation in the affected-side M1 ( r = 0.659, p = 0.006), further affirming the critical role of engaging the residual network on the affected side in facilitating functional recovery [ 51 ]. Phase synchronization analysis revealed that different training modes employ distinct strategies for regulating the reorganization of brain network connectivity. Following treatment, the assistive mode group demonstrated significant enhancement of intra-hemispheric frontal-motor network connectivity on the affected side, particularly by strengthening functional links between the ipsilesional PFC, DLPFC, and M1. These improvements may underpin a closed-loop control mechanism that incorporates motor planning, error monitoring, and real-time adjustments [ 52 – 54 ]. Conversely, the passive mode group showed increased connectivity within the cross-hemispheric sensory-visual pathway, such as between the contralesional SFC and the ipsilesional OC. This likely reflects a dependency on multisensory integration pathways but reveals a lack of improvement in direct motor-related connections, thereby constraining motor function recovery. Notably, the conventional group and the assistive mode group displayed similarities in cross-hemispheric motor network connectivity, such as links between the contralesional PMC and the ipsilesional M1, suggesting that both active participation training modes share certain mechanisms for cross-hemispheric information transmission. However, the assistive mode group, supported by high-precision, robotics-assisted repetitive training, appears to have further amplified these effects. At the neuroelectrophysiological level, while no statistically significant changes were observed in RMT, MEP latency, or MEP amplitude across the different groups, all groups exhibited a general trend of improvement in RMT. This finding suggests that short-term training may primarily enhance motor output function by improving synaptic efficacy plasticity, rather than inducing structural remodeling of the corticospinal tract. This conclusion aligns with the findings of Reis et al. regarding the underlying mechanisms of short-term rehabilitation interventions [ 55 ]. At the level of clinical application and translation, this study proposes a three-tier decision-making framework based on "injury severity–functional dimension–neural mechanisms." For patients with severely impaired baseline motor functions (FAC grade 0–1), assistive training is identified as the optimal approach, demonstrating clear clinical advantages by reconstructing the "motor intention–execution" pathway. For patients with mild to moderate functional impairments, equivalent training modes may be selected based on the allocation of medical resources. Furthermore, the study highlights dimension-specific effects of various training modes on functional recovery. Specifically, motor function (FMA) improvement is more reliant on prefrontal-motor network coactivation induced by assistive mode training, balance function (BBS) enhancement primarily depends on sensory input mechanisms bolstered by passive mode training, while progress in activities of daily living (MBI) is exclusively driven by active engagement modes [ 6 , 19 ]. Notably, a strong correlation was observed between near-infrared imaging features, such as enhanced prefrontal synchrony and interhemispheric connectivity in the occipital lobe, and functional prognosis. These findings provide a theoretical basis for the development of neuroimaging biomarkers and point to future applications, such as constructing machine learning models based on baseline brain activation patterns to enable the precise design of individualized rehabilitation plans. This study has several limitations. The small sample size (n = 44) may hinder the detection of certain inter-group effects, such as the main effect of FMA between groups ( p = 0.074), highlighting the need for larger multicenter studies to validate the robustness of the findings. Additionally, the two-week follow-up period restricted observation of long-term neuroplasticity effects. Future studies should consider extending this period to 3–6 months to evaluate functional maintenance effects and the dynamic processes of white matter remodeling. From a methodological perspective, the current spatial resolution of near-infrared technology limits the ability to analyze mechanisms in deeper brain regions. This necessitates integrating functional magnetic resonance imaging and diffusion tensor imaging techniques to further elucidate structure-function coupling patterns. Expanding the dimensionality of MEP parameter analysis (e.g., dynamic evaluation of motor threshold-output curves) is also recommended to detect subtle changes in pyramidal tract pathways. Future research should adopt a multidimensional approach by deeply integrating multimodal neuroimaging technologies, longitudinally tracking transition patterns between task-state and resting-state functional statuses, and incorporating neuroregulation methods such as transcranial magnetic stimulation and functional electrical stimulation. Establishing a tripartite precision rehabilitation system—"exoskeleton-brain-computer interface-neuroregulation"—could leverage synergistic mechanisms to facilitate neurocompensation and remodeling. Such advancements would offer robust, evidence-based support for personalized treatment, driving progress in rehabilitation science. Conclusion This study highlights that conventional therapies, robot-assisted gait training in passive mode, and robot-assisted gait training in assistive mode can significantly enhance motor function, balance ability, and daily living skills in stroke patients within a short timeframe. The varying activation patterns and neural network connections observed across different training modes suggest that these approaches may influence recovery through distinct neural mechanisms. Notably, assistive mode training demonstrates specific advantages in improving motor function and promoting neural activity reconstruction on the impaired side, whereas passive mode training uniquely contributes to enhancing balance ability and visual compensation pathways. Future research should incorporate advanced neuroimaging techniques and extended follow-up periods to examine the long-term impact of these rehabilitation strategies on brain remodeling and clinical outcomes. Such efforts would provide a more robust theoretical foundation for developing personalized rehabilitation plans. Abbreviations RAGT, Robot-assisted gait training; DLPFC, dorsolateral prefrontal cortex; M1, primary motor cortex; S1, primary somatosensory cortex; fNIRS, functional near-infrared spectroscopy; FMA-LE, Fugl-Meyer Assessment for the Lower Extremity; BBS, Berg Balance Scale; MBI, Modified Barthel Index; FAC, Functional Ambulation Category; RMT, resting motor threshold; MEP, motor-evoked potential; PFC, prefrontal cortex; SFC, superior frontal cortex; PMC, premotor cortex; OC, occipital cortex; HbO, oxygenated hemoglobin; PCA, principal component analysis; ICA, independent component analysis; WA, Wavelet Amplitude; WPCO, wavelet phase coherence; ANOVA, analysis of variance; LSD, least significant difference, SD, Standard Deviation; IQR, interquartile range; MD, mean difference; MCID, minimal clinically important difference Declarations Ethics approval and consent to participate This study was approved by the Medical Ethics Committee of the Third Affiliated Hospital of Sun Yat-sen University ([2021]02-333-03) in accordance with the declaration of Helsinki. All participants gave written informed consent prior to their enrollments. Consent for publication Consent from all authors has been acquired prior to submission of this article. Availability of data and materials The data that support the findings of this study are available from the corresponding author upon reasonable request. Competing interests The authors declare no competing interests. Funding Not applicable. Authors' contributions YJ.X. performed the conceptualization, investigation, drafted and wrote the main manuscript text. MS.S., XD.M., KX.C., and H.X. performed the experimental data acquisition, statistical analysis and prepared the figures. H.X. prepared the result visualization. ZL.D. is involved in project administration, as well as supervision and funding acquisition. ZL.D., and K.L were responsible for validation, reviewing, and editing the original work. All authors read and approved the final manuscript. Acknowledgements Not applicable Authors' information 1 Department of Rehabilitation Medicine, The Third Affiliated Hospital, Sun Yat-Sen University, Guangzhou, 510630, People’s Republic of China, 2 Guangzhou International Economics College, Guangzhou, 510450, People’s Republic of China 3 Department of Biomedical Engineering, Faculty of Engineering, Hong Kong Polytechnic University, Hong Kong, 999077, SAR, People’s Republic of China 4 Department of Hearing and Speech Science, Guangzhou Xinhua University, Guangzhou, 510520, People’s Republic of China References Campagnini S, et al. Effects of control strategies on gait in robot-assisted post-stroke lower limb rehabilitation: a systematic review. J Neuroeng Rehabil. 2022;19(1):52. Lin YN, et al. Hybrid robot-assisted gait training for motor function in subacute stroke: a single-blind randomized controlled trial. J Neuroeng Rehabil. 2022;19(1):99. Elmas Bodur B, Erdoğanoğlu Y, Asena S, Sel. Effects of robotic-assisted gait training on physical capacity, and quality of life among chronic stroke patients: A randomized controlled study. J Clin Neurosci. 2024;120:129–37. Leow XRG, Ng SLA, Lau Y. Overground Robotic Exoskeleton Training for Patients With Stroke on Walking-Related Outcomes: A Systematic Review and Meta-analysis of Randomized Controlled Trials. Arch Phys Med Rehabil. 2023;104(10):1698–710. Huang H, et al. Effect and optimal exercise prescription of robot-assisted gait training on lower extremity motor function in stroke patients: a network meta-analysis. Neurol Sci. 2025;46(3):1151–67. Huo C, et al. Effectiveness of unilateral lower-limb exoskeleton robot on balance and gait recovery and neuroplasticity in patients with subacute stroke: a randomized controlled trial. J Neuroeng Rehabil. 2024;21(1):213. Yang J, et al. Effect of robotic exoskeleton training on lower limb function, activity and participation in stroke patients: a systematic review and meta-analysis of randomized controlled trials. Front Neurol. 2024;15:1453781. Jiang Z, et al. Effects of body weight support training on balance and walking function in stroke patients: a systematic review and meta-analysis. Front Neurol. 2024;15:1413577. de Miguel-Fernández J, et al. Control strategies used in lower limb exoskeletons for gait rehabilitation after brain injury: a systematic review and analysis of clinical effectiveness. J Neuroeng Rehabil. 2023;20(1):23. Thimabut N, et al. Effects of the Robot-Assisted Gait Training Device Plus Physiotherapy in Improving Ambulatory Functions in Patients With Subacute Stroke With Hemiplegia: An Assessor-Blinded, Randomized Controlled Trial. Arch Phys Med Rehabil. 2022;103(5):843–50. Murata A, Wen W, Asama H. The body and objects represented in the ventral stream of the parieto-premotor network. Neurosci Res. 2016;104:4–15. Kleim JA, Jones TA. Principles of Experience-Dependent Neural Plasticity: Implications for Rehabilitation After Brain Damage. J Speech Lang Hear Res. 2008;51(1):S225–39. Song KJ, et al. The effect of robot-assisted gait training on cortical activation in stroke patients: A functional near-infrared spectroscopy study. NeuroRehabilitation. 2021;49(1):65–73. Peters S, et al. Passive, yet not inactive: robotic exoskeleton walking increases cortical activation dependent on task. J Neuroeng Rehabil. 2020;17(1):107. Bonnal J et al. Relation between Cortical Activation and Effort during Robot-Mediated Walking in Healthy People: A Functional Near-Infrared Spectroscopy Neuroimaging Study (fNIRS). Sens (Basel), 2022. 22(15). Louie DR, et al. Efficacy of an exoskeleton-based physical therapy program for non-ambulatory patients during subacute stroke rehabilitation: a randomized controlled trial. J Neuroeng Rehabil. 2021;18(1):149. Hirano S, et al. Effects of robot-assisted gait training using the Welwalk on gait independence for individuals with hemiparetic stroke: an assessor-blinded, multicenter randomized controlled trial. J Neuroeng Rehabil. 2024;21(1):76. Tam PK, et al. Overground robotic exoskeleton vs conventional therapy in inpatient stroke rehabilitation: results from a pragmatic, multicentre implementation programme. J Neuroeng Rehabil. 2025;22(1):3. Lee SY, et al. Effects of robot-assisted walking training on balance, motor function, and ADL depending on severity levels in stroke patients. Technol Health Care. 2024;32(5):3293–307. Schulz KF, Altman DG, Moher D. CONSORT 2010 statement: updated guidelines for reporting parallel group randomised trials. BMJ. 2010;340:c332. Louie DR, et al. Efficacy of an exoskeleton-based physical therapy program for non-ambulatory patients during subacute stroke rehabilitation: a randomized controlled trial. J Neuroeng Rehabil. 2021;18(1):149. Baker WB, et al. Modified Beer-Lambert law for blood flow. Biomed Opt Express. 2014;5(11):4053–75. Tan Q, et al. Frequency-specific functional connectivity revealed by wavelet-based coherence analysis in elderly subjects with cerebral infarction using NIRS method. Med Phys. 2015;42(9):5391–403. Pandian S, Arya KN, Kumar D. Minimal clinically important difference of the lower-extremity fugl-meyer assessment in chronic-stroke. Top Stroke Rehabil. 2016;23(4):233–9. Wu CY, et al. Constraint-induced therapy with trunk restraint for improving functional outcomes and trunk-arm control after stroke: a randomized controlled trial. Phys Ther. 2012;92(4):483–92. Taghavi A, Sharabiani P, et al. Minimal important difference of Berg Balance Scale, performance-oriented mobility assessment and dynamic gait index in chronic stroke survivors. J Stroke Cerebrovasc Dis. 2024;33(11):107930. Cramer SC. Repairing the human brain after stroke: I. Mechanisms of spontaneous recovery. Ann Neurol. 2008;63(3):272–87. Kleim JA, Jones TA. Principles of experience-dependent neural plasticity: implications for rehabilitation after brain damage. J Speech Lang Hear Res. 2008;51(1):S225–39. Pollock A et al. Interventions for improving upper limb function after stroke. Cochrane Database Syst Rev, 2014. 2014(11): p. Cd010820. Langhorne P, Coupar F, Pollock A. Motor recovery after stroke: a systematic review. Lancet Neurol. 2009;8(8):741–54. Winstein CJ, et al. Guidelines for Adult Stroke Rehabilitation and Recovery: A Guideline for Healthcare Professionals From the American Heart Association/American Stroke Association. Stroke. 2016;47(6):e98–169. Calautti C, Baron JC. Functional neuroimaging studies of motor recovery after stroke in adults: a review. Stroke. 2003;34(6):1553–66. Krakauer JW. Motor learning: its relevance to stroke recovery and neurorehabilitation. Curr Opin Neurol. 2006;19(1):84–90. Pascual-Leone A et al. THE PLASTIC HUMAN BRAIN CORTEX. Annual Review of Neuroscience, 2005. 28(Volume 28, 2005): pp. 377–401. Liepert J, et al. Treatment-Induced Cortical Reorganization After Stroke in Humans. Stroke. 2000;31(6):1210–6. Ward NS, et al. Motor system activation after subcortical stroke depends on corticospinal system integrity. Brain. 2006;129(3):809–19. Grefkes C, et al. Cortical connectivity after subcortical stroke assessed with functional magnetic resonance imaging. Ann Neurol. 2008;63(2):236–46. Moeller K, Willmes K, Klein E. A review on functional and structural brain connectivity in numerical cognition. Front Hum Neurosci, 2015. 9. Hampson M, et al. Brain Connectivity Related to Working Memory Performance. J Neurosci. 2006;26(51):13338–43. Mihara M, et al. Neurofeedback Using Real-Time Near-Infrared Spectroscopy Enhances Motor Imagery Related Cortical Activation. PLoS ONE. 2012;7(3):e32234. Badcock JC, Hugdahl K. A synthesis of evidence on inhibitory control and auditory hallucinations based on the Research Domain Criteria (RDoC) framework. Front Hum Neurosci, 2014. 8. Takeuchi N, et al. Repetitive Transcranial Magnetic Stimulation of Contralesional Primary Motor Cortex Improves Hand Function After Stroke. Stroke. 2005;36(12):2681–6. Murase N, et al. Influence of interhemispheric interactions on motor function in chronic stroke. Ann Neurol. 2004;55(3):400–9. Schaechter JD. Motor rehabilitation and brain plasticity after hemiparetic stroke. Prog Neurobiol. 2004;73(1):61–72. Kitago T, Krakauer JW. Chap. 8 - Motor learning principles for neurorehabilitation . In: Barnes MP, Good DC, editors. Handbook of Clinical Neurology. Editors: Elsevier; 2013. pp. 93–103. Kawai R, et al. Motor Cortex Is Required for Learning but Not for Executing a Motor Skill. Neuron. 2015;86(3):800–12. Gladstone DJ, Danells CJ, Black SE. The fugl-meyer assessment of motor recovery after stroke: a critical review of its measurement properties. Neurorehabil Neural Repair. 2002;16(3):232–40. Mahoney FI, Barthel DW. FUNCTIONAL EVALUATION: THE BARTHEL INDEX. Md State Med J. 1965;14:61–5. Lang CE, MacDonald JR, Gnip C. Counting Repetitions: An Observational Study of Outpatient Therapy for People with Hemiparesis Post-Stroke. J Neurol Phys Ther. 2007;31(1):3–10. Contreras-Vidal JL, et al. Neural Decoding of Robot-Assisted Gait During Rehabilitation After Stroke. Am J Phys Med Rehabil. 2018;97(8):541–50. Carter AR, et al. Resting interhemispheric functional magnetic resonance imaging connectivity predicts performance after stroke. Ann Neurol. 2010;67(3):365–75. Sitaram R, et al. Closed-loop brain training: the science of neurofeedback. Nat Rev Neurosci. 2017;18(2):86–100. Ridderinkhof KR, et al. The Role of the Medial Frontal Cortex in Cognitive Control. Science. 2004;306(5695):443–7. Hamada M, et al. The Role of Interneuron Networks in Driving Human Motor Cortical Plasticity. Cereb Cortex. 2012;23(7):1593–605. Reis J et al. Noninvasive cortical stimulation enhances motor skill acquisition over multiple days through an effect on consolidation. Proceedings of the National Academy of Sciences, 2009. 106(5): pp. 1590–1595. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6535268","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":452410727,"identity":"8942246d-78ec-44b8-ac33-01cb2b34e4fa","order_by":0,"name":"Yunjuan Xie","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABDUlEQVRIiWNgGAWjYDACZiCWqLCRY2BgbDwAE5QgrOVMmjFQSwORWkCAseVwYgOQJk6LwXHmYxKWDWnpa9sPA235c9je4ADzwds8DHZ5uLRINrOlSUjusMnddiax4QBj2+HEDQfYkq15GJKLcWnhZ+Yxk5A8k5a77QBIS8PhBIMDPGbSPAwHwE7FBtiY+b9JSLYdTjc7/xDmMP5veLUAbWEDaUkwuwG0hYHtMOOGAzxseLUA/WJsAQxkw203gLYktqUnzjzMZmw5xyAZpxaD84cf3gZGpbzZ+fSHDz78sbbnO9788MabCjucWoCARRoeDQkMzZDIZTDArR4ImD9+QHDq8CodBaNgFIyCkQkAeF5bx6zmtn4AAAAASUVORK5CYII=","orcid":"","institution":"Third Affiliated Hospital of Sun Yat-sen University","correspondingAuthor":true,"prefix":"","firstName":"Yunjuan","middleName":"","lastName":"Xie","suffix":""},{"id":452410728,"identity":"97e2a7c6-ceac-4f3d-bd32-3b0899cc2352","order_by":1,"name":"Meisi Song","email":"","orcid":"","institution":"Third Affiliated Hospital of Sun Yat-sen University","correspondingAuthor":false,"prefix":"","firstName":"Meisi","middleName":"","lastName":"Song","suffix":""},{"id":452410729,"identity":"af0e4968-6956-4e65-b933-2e34f9198a46","order_by":2,"name":"Xiaodan Ma","email":"","orcid":"","institution":"Third Affiliated Hospital of Sun Yat-sen University","correspondingAuthor":false,"prefix":"","firstName":"Xiaodan","middleName":"","lastName":"Ma","suffix":""},{"id":452410730,"identity":"5f371190-2cf7-4c9b-85bf-5bce512b42d8","order_by":3,"name":"Kaixuan Chen","email":"","orcid":"","institution":"Guangzhou International Economics College","correspondingAuthor":false,"prefix":"","firstName":"Kaixuan","middleName":"","lastName":"Chen","suffix":""},{"id":452410731,"identity":"2704df79-5394-4160-8d99-05f9bcf0b274","order_by":4,"name":"Hui Xie","email":"","orcid":"","institution":"Hong Kong Polytechnic University","correspondingAuthor":false,"prefix":"","firstName":"Hui","middleName":"","lastName":"Xie","suffix":""},{"id":452410732,"identity":"e8816142-fd0a-40f0-908c-427ad0bb2629","order_by":5,"name":"Kui Li","email":"","orcid":"","institution":"Third Affiliated Hospital of Sun Yat-sen University","correspondingAuthor":false,"prefix":"","firstName":"Kui","middleName":"","lastName":"Li","suffix":""},{"id":452410733,"identity":"2910c800-070e-4b7d-9147-e8e9f93f9709","order_by":6,"name":"Zulin Dou","email":"","orcid":"","institution":"Guangzhou Xinhua University","correspondingAuthor":false,"prefix":"","firstName":"Zulin","middleName":"","lastName":"Dou","suffix":""}],"badges":[],"createdAt":"2025-04-26 13:23:13","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6535268/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6535268/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":82309476,"identity":"ab063e91-a400-489c-adfc-0179157e42cb","added_by":"auto","created_at":"2025-05-09 01:44:34","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":44878,"visible":true,"origin":"","legend":"\u003cp\u003eOverview of the Bilateral Lower-Limb Exoskeleton System (AiWalker®)\u003c/p\u003e","description":"","filename":"Picture1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6535268/v1/46a28dea6b5babfecbec83c2.jpg"},{"id":82307721,"identity":"9133a21c-6eea-4d25-99c8-8b6a002668d2","added_by":"auto","created_at":"2025-05-09 01:28:33","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":12123,"visible":true,"origin":"","legend":"\u003cp\u003eLayout of Source and Detector Probes for fNIRS Measurements\u003c/p\u003e","description":"","filename":"Picture2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6535268/v1/abe8c5874ae23898a1e0f8ef.jpg"},{"id":82307723,"identity":"4a26e8f8-7860-4b35-bef7-8f320a6f69fa","added_by":"auto","created_at":"2025-05-09 01:28:34","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":37120,"visible":true,"origin":"","legend":"\u003cp\u003eCONSORT Flow Diagram of Patient Progression Through the Study\u003c/p\u003e","description":"","filename":"Picture3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6535268/v1/4eab81f78e3759665d602dc1.jpg"},{"id":82308738,"identity":"951478af-ae43-4845-9386-ef6da0195b79","added_by":"auto","created_at":"2025-05-09 01:36:34","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":85422,"visible":true,"origin":"","legend":"\u003cp\u003eComparative Analysis of Longitudinal Therapeutic Efficacy Among Conventional Rehabilitation, Passive and Assistive Mode Robotic Training\u003c/p\u003e","description":"","filename":"Picture4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6535268/v1/961753f94eaba9ef503ad115.jpg"},{"id":82307741,"identity":"f88d654b-9248-454e-a39b-a79f2e7304cf","added_by":"auto","created_at":"2025-05-09 01:28:34","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":125878,"visible":true,"origin":"","legend":"\u003cp\u003eNeural activation patterns across intervention groups pre- vs. post-intervention phases. (A-B: Assistive mode training group; C-D: Passive mode training group; E-F: Conventional therapies group). Warmer colors (red spectrum) indicate higher neural activation, while cooler hues (blue spectrum) represent lower activation levels.\u003c/p\u003e","description":"","filename":"Picture5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6535268/v1/6dee02e82e70c1ac7bd6808b.jpg"},{"id":82307732,"identity":"26e4d9d1-e0d4-4a8f-8575-17dc132d70aa","added_by":"auto","created_at":"2025-05-09 01:28:34","extension":"jpg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":178389,"visible":true,"origin":"","legend":"\u003cp\u003eComparative activation profiles of neural responses across therapeutic modalities. A/B, Assistive mode training (pre/post); C/D, Passive mode training (pre/post); E/F, Conventional therapy (pre/post). *p \u0026lt; 0.05, **p \u0026lt; 0.01, ***p \u0026lt; 0.001.\u003c/p\u003e","description":"","filename":"Picture6.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6535268/v1/e9c88613b976379b7e18909d.jpg"},{"id":82307738,"identity":"e132154f-6543-459f-b6c8-997ad58e3a78","added_by":"auto","created_at":"2025-05-09 01:28:34","extension":"jpg","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":136156,"visible":true,"origin":"","legend":"\u003cp\u003eInter-hemispheric (A) and intra-hemispheric (B) functional connectivity in conventional therapies vs. assistive mode training. Significant WPCO variations between (C) and within (D) hemispheres.\u003c/p\u003e","description":"","filename":"Picture7.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6535268/v1/169f90b95be0a57b9b091be4.jpg"},{"id":98774716,"identity":"07b1711a-b119-4ab5-8920-c92febadef0e","added_by":"auto","created_at":"2025-12-22 12:11:57","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1905416,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6535268/v1/0f9ec278-9f6e-4921-8803-abfe669e227a.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Comparative Effectiveness of Passive vs. Assistive Robotic Gait Training on Functional Recovery and Neuroplasticity Post-Stroke: A Randomized Controlled Trial","fulltext":[{"header":"Introduction","content":"\u003cp\u003eStroke-induced motor deficits in the lower extremities profoundly affect patients\u0026rsquo; ability to regain functional independence, participate in daily activities, and reintegrate into society. Consequently, gait rehabilitation constitutes a pivotal aspect of post-stroke recovery efforts [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Robot-assisted gait training (RAGT), utilizing innovations such as exoskeletons and end-effector systems, has emerged as a promising therapeutic approach, offering high-intensity and repetitive training. The efficacy of these robotic systems lies in their potential to enhance neuroplasticity, foster motor recovery, and improve mobility outcomes [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Evidence from systematic reviews and meta-analyses indicates that RAGT results in significant improvements in walking ability, speed, and balance, particularly when integrated with conventional therapies [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. However, the comparative clinical effectiveness of RAGT versus traditional rehabilitation methods, along with the fine-tuning of training protocols, continues to be a subject of active debate within stroke rehabilitation research [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eOver the past decade, significant progress has been made in developing robotic technologies and control strategies to optimize the interaction between patients and rehabilitation devices [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Assisted control strategies, which commonly involve trajectory tracking and compliant systems, are widely reported as prevalent methodologies in lower-limb exoskeletons [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. The efficacy of these strategies varies across distinct training modalities: assistive paradigms promote neural plasticity through active volitional engagement, while passive modalities primarily leverage sensory afferent mechanisms to facilitate cortical reorganization. To reinforce the research rationale, it is essential to explicitly contrast the neuroplasticity mechanisms of active engagement in assistive mode with sensory dependence in passive mode.\u003c/p\u003e \u003cp\u003eThe assistive mode requires patients to volitionally initiate movement intentions, engaging the dorsolateral prefrontal cortex (DLPFC) to encode motor goals. These goals subsequently modulate primary motor cortex (M1) activity via thalamocortical pathways [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. This process likely strengthens functional connectivity between the DLPFC and M1 through Hebbian plasticity, facilitating long-term cortical reorganization. In contrast, passive mode primarily relies on proprioceptive input, which activates the primary somatosensory cortex (S1) with minimal involvement of prefrontal cognitive processes [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Despite these distinct mechanisms, no research to date has directly compared the efficacy of assistive and passive mode training in terms of promoting neuroplasticity.\u003c/p\u003e \u003cp\u003eNeuroimaging and neurophysiological research have concurrently offered valuable insights into the neuroplasticity mechanisms activated by RAGT. Studies utilizing functional near-infrared spectroscopy (fNIRS) and other imaging modalities have demonstrated increased cortical activation in stroke patients after undergoing robotic training, particularly in the ipsilesional motor cortex and associated brain regions [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. The level of effort during robotic gait training has emerged as a pivotal factor, with evidence indicating that higher levels of effort during active walking activate broader cortical motor areas compared to fully assisted, passive walking in healthy individuals [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. While these findings are promising, further research is needed to clarify the precise role of neuroplasticity in functional recovery and its relationship to various robotic systems and control strategies. Discrepancies in the dose-response relationship\u0026mdash;specifically regarding training duration, frequency, and intensity\u0026mdash;add complexity to the development of standardized treatment protocols [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Moreover, although some robotic systems surpass conventional physiotherapy in specific areas, such as improving gait symmetry, endurance, and balance, the results are not universally consistent [\u003cspan additionalcitationids=\"CR17\" citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. The efficacy of RAGT also hinges on patient-specific factors, including the severity of motor impairment, ambulatory capacity, and the presence of comorbidities [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIn conclusion, this article aims to examine the therapeutic efficacy of various RAGT modes on stroke patients, with a concentrated emphasis on the recovery of lower limb motor function and the neural mechanisms underpinning this improvement. Additionally, it seeks to evaluate the impact of different training modes on brain neuroplasticity, including potential variations in brain activation and functional connectivity. The central hypothesis suggests that assistive mode training\u0026mdash;combining robotic assistance with the patient\u0026rsquo;s conscious effort to move\u0026mdash;outperforms passive training in delivering therapeutic benefits. This approach is expected to promote greater brain activation, enhanced functional connectivity between motor and sensory regions, and more significant cortical reorganization. By refining treatment protocols based on these findings, there is promise for maximizing functional recovery, minimizing long-term disability, and enhancing the quality of life for individuals recovering from stroke.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy Design and Participants\u003c/h2\u003e \u003cp\u003eThis study utilized a single-center, double-blind, randomized controlled trial design. A random allocation sequence was generated using a random number table, and patients were divided into three groups: the assistive mode training group, the passive mode training group, and the conventional rehabilitation therapy group. The grouping details were securely placed inside opaque envelopes. During the trial, these envelopes were opened sequentially based on the patients' enrollment order, with smaller random numbers corresponding to earlier enrolled participants. Patients were then assigned to either the experimental or control group according to the instructions contained within the envelopes. A double-blind approach was implemented, ensuring that both the patients and the outcome evaluators were blinded, while the intervention personnel remained unblinded.\u003c/p\u003e \u003cp\u003e With approval from the Ethics Committee of the Third Affiliated Hospital of Sun Yat-sen University, stroke patients with walking impairments hospitalized in the rehabilitation department of the same hospital and undergoing neurological rehabilitation training between September 2022 and May 2024 were recruited for the study. The trial was registered with the China Clinical Trial Registration Center under registration number ChiCTR2100054527. The study protocol adheres to the 2010 CONSORT guidelines for reporting parallel-group randomized trials [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe inclusion criteria specify patients aged 18 to 80 years who experienced their first unilateral stroke accompanied by walking dysfunction and provided informed consent, either personally or through their families. Exclusion criteria included individuals with non-stroke neurological conditions, such as craniocerebral trauma or neurocognitive disorders; those with cardiac pacemakers or intracranial metallic implants; patients with cranial defects or severe cognitive or communication impairments limiting emotional expression; individuals with a history of epilepsy or current pregnancy; and patients suffering from dysfunction of vital organs, including the heart, lungs, liver, or kidneys, or other severe systemic comorbidities incompatible with therapeutic interventions. All participants were fully briefed on the study's purpose and procedures and provided written informed consent before enrollment.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eExperimental Procedure\u003c/h3\u003e\n\u003cp\u003eEligible participants were stratified and randomly assigned to one of three therapeutic cohorts: the assistive mode group, which received robot-assisted gait training in assistive mode combined with conventional rehabilitation therapies; the passive mode group, which underwent robot-assisted gait training in passive mode alongside conventional therapies; and the control group, which followed traditional rehabilitation protocols exclusively. All groups participated in interventions of equal duration, monitored by blinded assessors to ensure consistency. Patients in both the assistive and passive mode groups received treatment for 30 minutes per day, five days a week, for two weeks, totaling 10 sessions.\u003c/p\u003e \u003cp\u003eThe RAGT was conducted using the bilateral lower-limb exoskeleton system (AiWalker\u0026reg;, manufactured by Beijing AI-Robotics Technology Co., Ltd., China) (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). This system integrates a closed-loop coordinated control mechanism to synchronize actuators at the bilateral hip and knee joints, ensuring alignment between the motor rotation centers and the anatomical axes of the corresponding joints. The exoskeleton is secured to the patient via adjustable straps positioned at the waist, thighs, calves, and feet. Key features include a fixed waist mechanism, powered lower-limb actuation, and support for overground walking. The waist support is equipped with adjustable movable linkages, enabling width customization to accommodate users with varying pelvic dimensions (measured as the distance between the left and right greater trochanters plus 2 cm).\u003c/p\u003e \u003cp\u003eThe system allows customization of hip and knee joint angles, offering maximal ranges of 33\u0026deg; flexion/23\u0026deg; extension for the hip and 53\u0026deg; flexion/full extension for the knee. Additionally, the gait cycle duration is adjustable, ranging between 2.45 and 5.25 seconds. The exoskeleton's anthropometric adaptability addresses multiple dimensions, including height, pelvic width, waist depth (defined as posterior protrusion from the midline), thigh length (measured from the greater trochanter to the knee joint gap plus 2 cm), and calf length (spanning the knee joint gap to the sole).\u003c/p\u003e \u003cp\u003eFor optimal use, patients must wear close-fitting clothing and flat-heeled shoes with heels shorter than 3 cm. Prior to usage, a pre-assessment of the lower-limb passive joint range of motion is mandatory.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo ensure proper donning, the patient is seated with their feet positioned on the footplates, ensuring that the heels align with the posterior edges, while leaning securely against the backplate. Straps are systematically tightened at the waist, thighs, calves, and feet to ensure proper fixation. In the passive mode training group, the exoskeleton delivers complete kinematic assistance for ambulation without requiring voluntary engagement of the lower limbs. The telescoping rods for the thighs and calves enable swift and precise length adjustments to accommodate various anthropometric variations. Clinician supervision is advised to ensure both safety and the effectiveness of the training process.\u003c/p\u003e \u003cp\u003eIn assistive mode training, the patient actively generates force to initiate movement, which is augmented by a motor aligned with the knee joint that provides additional torque to aid the completion of the action. Assistive mode training evaluates the patient\u0026rsquo;s force intensity over three distinct intervals: the first 10 minutes, the middle 15 minutes, and the final 5 minutes. Training quantification utilizes Newton-based metrics for subjective effort assessment, supported by real-time sensor data acquisition. Clearly defined participation criteria require dynamic torque modulation to maintain a minimum of 60% volitional force contribution throughout the training sessions.\u003c/p\u003e \u003cp\u003eAll participants underwent standard inpatient rehabilitation during hospitalization, which included neurodevelopmental treatment based on the Bobath concept. This therapy was administered exclusively by certified physiotherapists accredited by the International Bobath Instructors Training Association. The core interventions were as follows: (1) Key point control to enhance dynamic stabilization (e.g., pelvis and scapular girdle stabilization); (2) Postural preparation training, such as trunk rotation with weight shifting while sitting and contralesional stepping during weight-bearing on the paretic limb; (3) Task-oriented exercises, including bridging drills to activate the gluteus maximus and transverse abdominis, sit-to-stand transfers facilitated by tactile cues (e.g., therapist\u0026rsquo;s hand placement on the knees to promote eccentric quadriceps control), and balance training on unstable surfaces coupled with visual distractions (e.g., ball-catching exercises); and (4) Gait re-education featuring body-weight support harnesses to refine segmented gait phases (e.g., heel strike, knee extension, and hip extension timing), paired with mirror feedback. Each session lasted 50 minutes and was conducted daily over a two-week period (10 sessions in total). The training parameters, such as surface stability and resistance levels, were customized to align with each patient\u0026rsquo;s functional capacity, following the methodology outlined by Louie et al. [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e].\u003c/p\u003e\n\u003ch3\u003eClinical Assess\u003c/h3\u003e\n\u003cp\u003eBaseline data were collected prior to the initiation of the trial, capturing demographic variables (age, gender, height, weight, dominant hand) and clinical characteristics (lesion location, stroke type, time since stroke onset, stroke severity, and pre-intervention functional status). The primary outcome measure focused on lower limb motor recovery, assessed using the Fugl-Meyer Assessment for the Lower Extremity (FMA-LE). Secondary outcome measures included the Berg Balance Scale (BBS), Modified Barthel Index (MBI), Functional Ambulation Category (FAC), and corticospinal excitability parameters, such as resting motor threshold (RMT), motor-evoked potential (MEP) latency, and MEP amplitude. Blinded assessments were carried out at two time points: pre-intervention (T0) and post-2-week intervention (T1). These evaluations were performed by two independent physical therapists who had undergone a standardized assessor training program to ensure interrater reliability. Additionally, safety monitoring was conducted throughout the trial to document any adverse events.\u003c/p\u003e\n\u003ch3\u003efNIRS Data Acquisition and Processing\u003c/h3\u003e\n\u003cp\u003efNIRS data acquisition was performed using the Nirsmart system (Danyang Huichuang Medical Equipment Co., Ltd., China), which is equipped with 18 emitter probes transmitting dual wavelengths (740 nm and 850 nm) and 16 detector probes arranged with a standardized source-detector separation of 30 mm. Optode placement adhered to the 10\u0026ndash;20 international system, targeting specific brain regions, including the bilateral M1, S1, DLPFC, prefrontal cortex (PFC), superior frontal cortex (SFC), premotor cortex (PMC), and occipital cortex (OC) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). During operation, ensure that the optodes are vertically and securely attached to the scalp, with optical gel applied for participants with long hair to minimize refraction loss.\u003c/p\u003e \u003cp\u003eThe experimental environment should be maintained at a stable temperature (20\u0026ndash;25\u0026deg;C) and free from light interference, with participants using a head support frame to reduce motion artifacts. Continuous signal recording was conducted at a sampling frequency of 10 Hz. Before data collection, a 10-minute resting-state baseline calibration was performed, with real-time monitoring of the signal quality index. Channels with poor signal quality were adjusted or excluded as necessary. After the experiment, the optodes should be thoroughly cleaned, and the intensity of the light source should be periodically inspected.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe raw optical intensity signals are converted into oxygenated hemoglobin (HbO) concentrations using a modified Beer-Lambert law [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Motion artifacts are addressed through the following steps: (1) artifact detection based on a moving standard deviation, identifying deviations exceeding 10% of the baseline; (2) cubic spline interpolation to correct transient artifacts lasting less than 2 seconds; (3) noise removal using principal component analysis (PCA), where the top three components accounting for over 95% of the variance are extracted; and (4) physiological noise reduction implemented with Butterworth bandpass filtering (0.01\u0026ndash;0.5 Hz) and independent component analysis (ICA). After segmentation, the data undergoes baseline normalization using z-score standardization. Subsequently, the HbO signals are analyzed in the time-frequency domain through Morlet wavelet transformation within a 4\u0026ndash;60-second window to isolate 0.01\u0026ndash;0.08 Hz band power during task phases.\u003c/p\u003e \u003cp\u003eWavelet Amplitude (WA) represents a precise quantitative metric for assessing cortical activation intensity. The computational process unfolds as follows: Pre-processed time-domain signals obtained from fNIRS recordings undergo continuous wavelet transformation, generating time-frequency representations. This approach facilitates the temporal integration of frequency-specific components, producing time-resolved WA values. In essence, WA reflects the magnitude of task-induced hemoglobin oscillatory amplitudes within the cortical microvasculature, where higher WA values directly indicate enhanced neurovascular coupling responses in specific cerebral regions during task execution. Notably, robust WA measurements signify heightened functional specificity between cortical activation patterns and the cognitive-motor demands at hand.\u003c/p\u003e \u003cp\u003eFunctional connectivity has long served as a cornerstone metric for exploring inter-regional cerebral interactions, with its innovative methodological approach epitomized by the wavelet phase coherence (WPCO) algorithm [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. Central to this framework is the decomposition of raw signals into time-frequency representations via continuous wavelet transform, which facilitates the extraction of instantaneous phase information across oscillatory frequency bands. The analytical pipeline unfolds across three distinct tiers: First, dynamic phase sequences are constructed from time-varying frequency components. Subsequently, the temporal coherence of inter-signal phase differences is quantified to yield WPCO coefficients\u0026mdash;a parametric measure of cross-regional rhythmic synchronization, where higher values signify intensified neural coordination. In the final tier, an amplitude-adaptive Fourier transform technique is employed to mitigate amplitude-mediated phase distortion. This step effectively eliminates artifactual connectivity, isolating neurophysiologically valid functional coupling strengths that accurately reflect genuine neural activity dynamics.\u003c/p\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003e This study employed a three-arm randomized controlled trial with a repeated measures design, featuring two assessment points (pre- and post-intervention) for each participant across all groups. The sample size was determined using G*Power 3.1 software to analyze the group-by-time interaction effect, which constituted the primary hypothesis. Calculations were based on the assumption of a large effect size (Cohen\u0026rsquo;s \u003cem\u003ef\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.6), a two-tailed type I error rate (\u003cem\u003eα\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.05), a type II error rate (\u003cem\u003eβ\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.20) corresponding to a statistical power of 0.8, a three-group structure, two repeated measurements, an assumed correlation coefficient of 0.5 among repeated measures, and a non-sphericity correction factor (\u003cem\u003eε\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1.0) under the assumption of sphericity. The results indicated that a minimum of 33 participants (11 per group) would be required. To accommodate an anticipated attrition rate of 10%, the final sample size was increased to 39 participants, with 13 individuals assigned to each group.\u003c/p\u003e \u003cp\u003eThis trial followed the Per-Protocol principle, focusing exclusively on patients who strictly adhered to the study protocol. These patients met the study design criteria, received all prescribed treatments, and completed follow-up as outlined. The normality of the data was evaluated using the D'Agostino-Pearson test. Quantitative data with a normal distribution were reported as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation (\u003cem\u003ex̄\u003c/em\u003e \u0026plusmn; \u003cem\u003es\u003c/em\u003e), while non-normally distributed data were presented as the median (\u003cem\u003eP\u003c/em\u003e\u003csub\u003e\u003cem\u003e25\u003c/em\u003e\u003c/sub\u003e, \u003cem\u003eP\u003c/em\u003e\u003csub\u003e\u003cem\u003e75\u003c/em\u003e\u003c/sub\u003e). For baseline comparisons, independent t-tests were conducted for continuous variables, and \u003cem\u003eχ\u0026sup2;\u003c/em\u003e tests were used for categorical variables.\u003c/p\u003e \u003cp\u003eOutcome analyses of normally distributed quantitative data included homogeneity of variance and sphericity assessments. When assumptions were satisfied, repeated-measures analysis of variance (ANOVA) was performed, with treatment groups (passive, assistive, and conventional) as between-subject factors and measurement times (T0, T1) as within-subject factors. Statistically significant results were further examined using least significant difference (LSD) post-hoc tests. In cases where sphericity assumptions were violated, the Greenhouse-Geisser correction was applied. For non-normally distributed or heteroscedastic data, comparisons between groups were conducted using the Kruskal-Wallis H test, while within-group comparisons utilized Wilcoxon signed-rank tests.\u003c/p\u003e \u003cp\u003eStatistical significance was defined as two-tailed \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05. Adverse events were analyzed to determine their relevance to the trial and evaluate overall safety. Data processing and analysis were performed using SPSS 22.0 software.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eParticipant Flow and Characteristics\u003c/h2\u003e \u003cp\u003eA total of 48 patients, meeting the inclusion criteria, were recruited from the inpatient department of The Third Affiliated Hospital of Sun Yat-sen University between September 2022 and May 2024. Participants were randomly assigned in a 1:1:1 ratio to one of three groups: the passive mode training group, the assistive mode training group, or the conventional therapy group. Four patients in the passive mode training group withdrew for personal reasons after the initial assessment but prior to the start of treatment. Consequently, 44 patients completed the two-week intervention and underwent outcome evaluations (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eMost participants had experienced a disease duration of 1 to 6 months. No significant differences were noted among the groups with respect to demographic variables (age, gender, height, weight, dominant hand) or clinical characteristics, including lesion location, stroke type, time since stroke onset, stroke severity, and pre-intervention functional status (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Additionally, baseline outcome measurements revealed no intergroup differences (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e), ensuring data comparability. No adverse events were reported throughout the study.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eBaseline Characteristics\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAssistive mode training group\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;16)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePassive mode training group\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;12)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eConventional therapies group\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;16)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAge, mean (SD), year\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e58.44 (8. 15)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e58.11 (8.25)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e56.31 (12.66)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMale patients, No. (%)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e11(68.75%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6(50%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e10(62.50%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eWeight, median (IQR), kg\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e65(55.25\u0026ndash;68.75)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e65(46.50\u0026ndash;70.70)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e61(56.18\u0026ndash;66.80)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eHeight, mean (SD), cm\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e168.00 (7.69)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e164.11 (8.64)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e161.25(5.60)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eBody mass index\u003c/b\u003e,\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e21.94(3.13)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e22.35(4.54)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e23.88(3.73)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eIschemic, No. (%)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e10(62.50%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7(58.33%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e9(56.25%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAffected side, No. (%)\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u003cb\u003eLeft\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u003cb\u003eRight\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e8(50%)\u003c/p\u003e \u003cp\u003e8(50%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5(41.67%)\u003c/p\u003e \u003cp\u003e7(58.33%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e10(62.50%)\u003c/p\u003e \u003cp\u003e6(37.50%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTime from stroke, median (IQR), days\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e96(41-139.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e42(31\u0026ndash;86)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e95(51.25\u0026ndash;166.5)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eSD, Standard Deviation; IQR, interquartile range.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eLongitudinal Trajectories of Neuromotor Performance Metrics and Functional Recovery Indices in Robotic-Assisted vs. Conventional Therapy\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAssistive mode training group\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;16)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePassive mode training group\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;12)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eConventional therapies group\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;16)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eBetween-group difference (\u003cem\u003ep\u003c/em\u003e)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFMA-LE (score)\u003c/p\u003e \u003cp\u003eT0\u003c/p\u003e \u003cp\u003eT1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e19.02 (5.32)\u003c/p\u003e \u003cp\u003e20.81(5.26)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e13.33 (6.44)\u003c/p\u003e \u003cp\u003e17.00(7.47)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e20.00 (7.98)\u003c/p\u003e \u003cp\u003e22.69(6.26)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.059\u003c/p\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eWithin-group difference\u003c/b\u003e (\u003cem\u003ep\u003c/em\u003e/95%CI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.039(0.107 to 3.471)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;0.001(1.991 to 5.342)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.005(0.950 to 4.425)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eBBS(score)\u003c/b\u003e\u003c/p\u003e \u003cp\u003eT0\u003c/p\u003e \u003cp\u003eT1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e29.38 (10.54)\u003c/p\u003e \u003cp\u003e34.70(10.87)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e19.22 (16.07)\u003c/p\u003e \u003cp\u003e25.78(17.61)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e31.50 (14.99)\u003c/p\u003e \u003cp\u003e37.19(13.10)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.099\u003c/p\u003e \u003cp\u003e0.129\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eWithin-group difference\u003c/b\u003e (\u003cem\u003ep\u003c/em\u003e/95%CI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;0.001(2.995 to 7.645)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.005(2.559 to 10.552)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;0.001(3.050 to 8.325)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMBI(score)\u003c/b\u003e\u003c/p\u003e \u003cp\u003eT0\u003c/p\u003e \u003cp\u003eT1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e51.69(22.69)\u003c/p\u003e \u003cp\u003e63.19(21.38)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e47.89(22.70)\u003c/p\u003e \u003cp\u003e53.44(21.86)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e61.75(23.07)\u003c/p\u003e \u003cp\u003e67.69(19.83)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.284\u003c/p\u003e \u003cp\u003e0.273\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eWithin-group difference\u003c/b\u003e (\u003cem\u003ep\u003c/em\u003e/95%CI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;0.001(5.624 to 17.376)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.088(-1.036 to 12.147)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.005(2.039 to 9.836)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eFAC(score), median (IQR)\u003c/b\u003e\u003c/p\u003e \u003cp\u003eT0\u003c/p\u003e \u003cp\u003eT1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.5(0\u0026thinsp;~\u0026thinsp;3)\u003c/p\u003e \u003cp\u003e2(0.25\u0026thinsp;~\u0026thinsp;3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1(0\u0026thinsp;~\u0026thinsp;1.5)\u003c/p\u003e \u003cp\u003e1(0\u0026thinsp;~\u0026thinsp;2.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.5(0\u0026thinsp;~\u0026thinsp;3)\u003c/p\u003e \u003cp\u003e2(0.25\u0026thinsp;~\u0026thinsp;3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.714\u003c/p\u003e \u003cp\u003e0.528\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eWithin-group difference\u003c/b\u003e (\u003cem\u003ep\u003c/em\u003e/95%CI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.015\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.081\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.014\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eRMT (%)\u003c/b\u003e\u003c/p\u003e \u003cp\u003eT0\u003c/p\u003e \u003cp\u003eT1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e24.53 (9.17)\u003c/p\u003e \u003cp\u003e23.80(10.14)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e27.44 (5.25)\u003c/p\u003e \u003cp\u003e25.11(7.80)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e25.63 (10.09)\u003c/p\u003e \u003cp\u003e24.69(9.37)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.741\u003c/p\u003e \u003cp\u003e0.938\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eWithin-group difference\u003c/b\u003e (\u003cem\u003ep\u003c/em\u003e/95%CI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.350(-2.361 to 0.894)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.180(-6.00 to 1.333)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.649(-5.242 to 3.367)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMEP latency (ms)\u003c/b\u003e\u003c/p\u003e \u003cp\u003eT0\u003c/p\u003e \u003cp\u003eT1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e34.19 (5.43)\u003c/p\u003e \u003cp\u003e35.39(5.24)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e34.46 (3.09)\u003c/p\u003e \u003cp\u003e30.72(4.21)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e32.92(4.91)\u003c/p\u003e \u003cp\u003e31.26(5.34)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.674\u003c/p\u003e \u003cp\u003e\u003cb\u003e0.035\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eWithin-group difference\u003c/b\u003e (\u003cem\u003ep\u003c/em\u003e/95%CI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.241(-0.905 to 3.313)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.126(-8.800 to 1.309)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.346(-5.316 to 1.983)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMEP amplitude (\u0026micro;V)\u003c/b\u003e\u003c/p\u003e \u003cp\u003eT0\u003c/p\u003e \u003cp\u003eT1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e239.70 (90.09)\u003c/p\u003e \u003cp\u003e219.19(77.95)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e195.58 (75.02)\u003c/p\u003e \u003cp\u003e258.27(127.51)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e187.62 (61.64)\u003c/p\u003e \u003cp\u003e214.21(100.88)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.152\u003c/p\u003e \u003cp\u003e0.544\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eWithin-group difference\u003c/b\u003e (\u003cem\u003ep\u003c/em\u003e/95%CI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.506(-84.974 to 43.954)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.209(-43.136 to 168.507)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.331(-29.804 to 82.979)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eFMA-LE, Fugl-Meyer Assessment for Lower Extremity; BBS, Berg Balance Scale; MBI, Modified Barthel Index; FAC, Functional Ambulatory Category; RMT, resting motor threshold; MEP, motor evoked potential.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eClinical Outcomes and Assessments\u003c/h3\u003e\n\u003cp\u003eData on lower extremity motor performance, balance, gait patterns, and functional status parameters in daily living are visually represented in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, with corresponding numerical analyses detailed comprehensively in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. Repeated-measures ANOVA results for FMA-LE revealed the following: Mauchly\u0026rsquo;s test identified a violation of sphericity, necessitating Greenhouse-Geisser correction for degrees of freedom. No statistically significant time \u0026times; group interaction was observed (p\u0026thinsp;=\u0026thinsp;0.331), indicating consistent improvement trends across all three groups over time. Regarding main effects, a significant time effect was identified (F(1, 41)\u0026thinsp;=\u0026thinsp;30.761, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001, partial \u003cem\u003eη\u0026sup2;\u003c/em\u003e = 0.447), highlighting substantial improvement in patient scores across all groups. Post-treatment scores revealed significant increases in the conventional therapies group [mean difference (MD)\u0026thinsp;=\u0026thinsp;2.688, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.005)] and the passive mode training group (MD\u0026thinsp;=\u0026thinsp;3.667, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), while the assistive mode training group demonstrated a highly significant effect (MD\u0026thinsp;=\u0026thinsp;1.789, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.039). A marginally significant between-group main effect was also noted (F(2, 41)\u0026thinsp;=\u0026thinsp;2.874, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.069, partial \u003cem\u003eη\u0026sup2;\u003c/em\u003e = 0.131), which may be attributed to the limited sample size, warranting further validation with larger study cohorts. Importantly, the observed within-group difference of 6.1 points in FMA-LE surpasses the established minimal clinically important difference (MCID) threshold of 6 points for chronic stroke [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e], underscoring the clinically meaningful advantage of robotic assistive mode training.\u003c/p\u003e \u003cp\u003eThe repeated-measures ANOVA results for BBS scores across the three groups yielded the following observations: At both T0 and T1, all groups showed improvement in mean BBS scores over time. Post-correction analysis revealed no significant time \u0026times; group interaction (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.825), indicating parallel improvement trajectories among the groups. A statistically significant main effect of time was identified (F(1, 41)\u0026thinsp;=\u0026thinsp;57.110, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001, partial \u003cem\u003eη\u0026sup2;\u003c/em\u003e = 0.600), with substantial post-intervention score increases observed across all groups: conventional therapy group (MD\u0026thinsp;=\u0026thinsp;5.688, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), passive mode training group (MD\u0026thinsp;=\u0026thinsp;6.556, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.005), and assistive mode training group (MD\u0026thinsp;=\u0026thinsp;5.320, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Notably, the passive mode training group demonstrated the most pronounced improvement. No statistically significant between-group main effect was detected (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.106), suggesting comparable efficacy among the intervention strategies at the analyzed time points. Furthermore, the inter-group BBS improvement exceeded the chronic stroke MCID threshold (\u0026ge;\u0026thinsp;3.2 points), indicating measurable clinical effects across all groups. These findings support the implementation of each intervention as an evidence-based balance recovery strategy [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe time \u0026times; group interaction for MBI showed no statistical significance (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.167), indicating that the improvement trajectories over time were comparable across the three groups. However, a significant main effect of time was observed (F(1,41)\u0026thinsp;=\u0026thinsp;26.464, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001, partial \u003cem\u003eη\u0026sup2;\u003c/em\u003e = 0.411), demonstrating substantial clinical score improvements across all participants. Post-hoc analyses revealed statistically significant improvements in the conventional group (MD\u0026thinsp;=\u0026thinsp;5.938, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.005) and the assistive mode group (MD\u0026thinsp;=\u0026thinsp;11.5, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.0001), while the passive mode group exhibited no significant change (MD\u0026thinsp;=\u0026thinsp;5.556, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.088). Furthermore, no significant differences between the groups were identified (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.286).\u003c/p\u003e \u003cp\u003eAt T0, no statistically significant differences in FAC scores were observed among the three groups (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.714). Similarly, no notable between-group differences were detected in FAC scores at T1 (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.528). Within-group comparisons showed that changes in FAC scores within each group did not achieve statistical significance. However, a stratified analysis based on baseline FAC scores revealed statistically significant differences among the three groups within the FAC levels 0\u0026ndash;1 category. Notably, a clinically meaningful distinction emerged between the assistive mode group and the passive mode group, with a mean difference of 5.49 (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.029; 95% CI: 0.61 to 10.37).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eCorticospinal Excitability Parameters\u003c/h2\u003e \u003cp\u003eThe repeated-measures ANOVA for RMT indicated preserved sphericity (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05), showing no statistically significant time \u0026times; group interaction (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.782) or between-group differences (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.602). Modest RMT improvements were noted across all groups at the two-week follow-up; however, intergroup variations remained non-significant (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.169). In the MEP latency analysis, the findings revealed non-significant between-group differences (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.147) and no notable temporal effects (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.149). Regarding MEP amplitude, no substantial improvements were observed in any group. The analysis confirmed compliance with sphericity (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05), with non-significant time \u0026times; group interaction (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.236), negligible between-group differences (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.446), and undetectable temporal variations (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.239).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eCortical Activation and Functional Connectivity\u003c/h2\u003e \u003cp\u003eAfter two weeks of varied training modes, distinct activation patterns in task-related brain regions emerged across the groups. A horizontal comparison between the conventional therapy group and the assistive mode training group revealed significant differences in task-rest activation values within the following brain regions during task execution: the contralesional SFC (t\u0026thinsp;=\u0026thinsp;2.407, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.022), the contralesional OC (t\u0026thinsp;=\u0026thinsp;2.169, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.038), the ipsilesional M1 (t\u0026thinsp;=\u0026thinsp;2.383, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.024), and the ipsilesional S1 (t\u0026thinsp;=\u0026thinsp;2.354, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.025). These differences remained statistically significant even after applying Bonferroni correction. Conversely, no statistically significant activation differences were observed between the conventional therapy group and the passive mode group or between the assistive mode group and the passive mode group (p\u0026thinsp;\u0026gt;\u0026thinsp;0.05).\u003c/p\u003e \u003cp\u003eLongitudinal analysis revealed that patients in the assistive mode training group demonstrated notable neuroplasticity in task-related activation patterns following two weeks of RAGT. Specifically, activation intensity under assistive mode conditions showed significant changes from baseline in the bilateral SFC, bilateral PMC, the contralesional S1, and the ipsilesional M1 (all \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05). In contrast, the control group and passive intervention group displayed no significant pre- and post-training changes in activation values during identical task-state conditions.\u003c/p\u003e \u003cp\u003ePrior to the intervention, cross-phase analysis of neural activation patterns revealed a dominant activation within the ipsilesional hemisphere during task conditions for all patients (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). Post-intervention assessments in the conventional group showed reduced activation across bilateral cortical networks under task-elicited conditions. Neuroimaging analyses after treatment identified a predominant activation pattern localized to the ipsilesional cerebral hemisphere (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05). In contrast, the assistive mode group displayed notable neuro-rebalancing effects, characterized by a significant reduction in contralesional hemisphere activation compared to pre-training, alongside an increased proportion of activation in the primary functional areas of the ipsilesional hemisphere. Meanwhile, the passive mode group did not achieve statistically significant differences in task-rest activation across either hemisphere following training (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe characteristics of cortical functional network reorganization in stroke patients under different training modes are summarized as follows: When comparing the conventional group to the assistive mode group, significant differences were identified in interhemispheric functional connectivity. These differences specifically involve the contralesional PMC and ipsilesional M1 (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.036); the ipsilesional PFC and contralesional PMC (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.025); as well as the ipsilesional PFC and contralesional S1 (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.048).\u003c/p\u003e \u003cp\u003eIn terms of intra-hemispheric functional networks, statistical differences were observed between the two groups within the contralesional hemisphere, particularly in the PFC-SFC (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.037), PFC-S1 (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.024), and PFC-OC (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.042) connectivities. Similarly, differences were observed within the ipsilesional hemisphere, involving the SFC-S1 (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.027) and PMC-S1 (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.026) connectivities.\u003c/p\u003e \u003cp\u003eFollowing assistive mode training, patients demonstrated neuroplastic changes. Notably, intra-hemispheric functional connectivity in the contralesional hemisphere increased significantly compared to baseline. Newly formed functional couplings were observed between the S1 and SFC (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05), S1 and PMC (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05), as well as S1 and M1 (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05). In the ipsilesional hemisphere, statistically significant functional connectivity was noted between the PFC and DLPFC, SFC and PMC, SFC and S1, S1 and M1, and S1 and OC (all \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05).\u003c/p\u003e \u003cp\u003eLateral comparison analysis between the conventional group and the passive mode group highlighted inter-group differences in interhemispheric functional connectivity, specifically between the contralesional PFC and ipsilesional DLPFC (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.045). After two weeks of passive training, patients in the passive mode group exhibited a distinct pattern of reorganization. Changes in interhemispheric connectivity were focused between the contralesional SFC and ipsilesional OC, as well as the contralesional S1 and ipsilesional M1. Intra-hemispheric connectivity changes were observed between the contralesional DLPFC and SFC, contralesional SFC and S1, ipsilesional SFC and PMC, and ipsilesional S1 and M1 (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eAssociations Between Motor Improvements and Neurophysiological Metrics\u003c/h2\u003e \u003cp\u003eThe analysis of neurofunctional score\u0026ndash;brain activation correlations revealed distinct patterns across groups. In the conventional group, higher MBI scores showed strong positive associations with increased activation in the contralesional S1 (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.785, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), ipsilesional SFC (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.823, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and ipsilesional M1 (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.659, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.006), reflecting large and moderate effect sizes, respectively. In contrast, the assistive mode group demonstrated significant MBI improvements accompanied by heightened neural synchronization in the ipsilesional PFC (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.580, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.019), highlighting the role of prefrontal regulatory mechanisms in motor recovery. Meanwhile, the passive mode group exhibited a positive correlation between functional gains and contralesional OC activation (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.747, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.021), pointing to the involvement of visual processing networks in compensatory strategies.\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study systematically examines the differences in the roles and neural mechanisms of passive mode, assistive mode, and conventional rehabilitation training in the motor function recovery of stroke patients. It utilizes an integrative approach involving behavioral assessments (FMA-LE, BBS, MBI, FAC), neuroelectrophysiology (MEP), and functional imaging (WA, WPCO). The findings indicate that all training groups achieved sustained functional improvements over time, consistent with the \"time-sensitive window\" phenomenon of post-stroke neuroplasticity [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eSignificantly, variations in the efficacy of functional improvements were observed across the different training modes. While the passive mode group exhibited the greatest improvement in balance function, as reflected in BBS scores (MD\u0026thinsp;=\u0026thinsp;6.556), the intergroup difference was not statistically significant (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05). This suggests that balance function recovery may depend more on consistent training intensity rather than the specific training mode employed [\u003cspan additionalcitationids=\"CR30\" citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. The observed effect is likely attributable to sustained proprioceptive input in the passive mode, which may enhance postural control abilities through the activation of the cerebellum-vestibular pathway [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eConversely, the assistive mode demonstrated distinct advantages in both balance function (BBS) and activities of daily living (MBI). Notably, MBI scores showed a highly significant increase (MD\u0026thinsp;=\u0026thinsp;11.5, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.0001), underscoring the critical importance of active participation in maximizing functional gains. This improvement is achieved through the reinforcement of closed-loop regulation within the motor planning and execution system [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. For instance, prior studies have indicated that active training can induce motor cortex reorganization, providing an optimal neural environment for motor rehabilitation [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe FAC-stratified analysis revealed differences in treatment responses; however, these findings should be interpreted cautiously given the limited statistical power. The post hoc nature of the analysis, coupled with small sample sizes within each stratum, increases the risk of Type II errors, positioning these results as hypothesis-generating rather than definitive. The stratification of treatment efficacy by baseline injury severity offers deeper insights into the mechanisms of action of various training modalities. In the subgroup with severe baseline motor dysfunction (FAC grades 0\u0026ndash;1), the assistive mode group demonstrated significantly greater FMA improvement compared to the passive mode group, offering valuable guidance for clinical decision-making.\u003c/p\u003e \u003cp\u003eNeuroimaging evidence indicates that patients with severe dysfunction often struggle to voluntarily initiate effective movements due to disrupted motor network connectivity on the affected side [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]. Assistive training mitigates this challenge by reducing task difficulty through real-time robotic assistance, enabling patients to attempt active movements within a manageable range. This fosters a positive cycle of \u0026ldquo;task difficulty adaptation\u0026ndash;affected-side network activation\u0026ndash;synaptic efficacy enhancement,\u0026rdquo; ultimately helping to re-establish neural pathways between motor intention and execution [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. This proposed mechanism is supported by near-infrared imaging, which shows enhanced functional connectivity between the PFC and M1 on the affected side in the assistive mode group [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e]. Additionally, prior research suggests that compensatory activation of the unaffected side may inhibit the remodeling potential of the affected side [\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eConversely, passive mode training proves less effective in the subgroup with severe dysfunction, as its reliance on external drive hinders cortical active reorganization [\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. Nonetheless, passive mode training retains value; through continuous mechanical assistance, it generates rhythmic joint movements that deliver substantial proprioceptive and tactile input to the central nervous system [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e]. These sensory signals, transmitted via the thalamocortical pathways to premotor and sensory integration areas, play a critical role in refining motor planning and enhancing muscle coordination [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e]. Even in the absence of active participation, patients in the passive mode group in this study improved balance control through visual feedback mechanisms. This effect may be attributed to compensatory activation of the occipital cortex on the unaffected side. As Krakauer (2006) noted, the visual system plays a pivotal role in motor control compensation, particularly when motor execution is impaired. Visual information can help reorganize motor function by providing spatial positioning cues [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eOverall, these findings highlight the necessity for clinical rehabilitation practices to consider the distinct impacts of different training modalities on the sensorimotor integration system [\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eFrom the perspective of neural remodeling mechanisms, fNIRS demonstrates that the brain\u0026rsquo;s network reconstruction, induced by different training modes, displays distinct paradigm-dependent characteristics. Patients in the assistive mode group exhibited a pronounced affected-side dominant activation pattern following training, marked by significant activation (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) in the bilateral SFC, PMC, and affected-side M1 during task states [\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e]. Moreover, their improvement in MBI showed a significantly positive correlation with increased neural synchrony in the ipsilesional PFC. This observation aligns closely with the task-specific neural remodeling theory [\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e], suggesting that the assistive mode enhances functional coupling between prefrontal regulatory regions and motor execution areas, thereby facilitating the efficient conversion of motor intentions into actual movements [\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIn contrast, the passive mode group demonstrated a unique unaffected-side compensatory mechanism. In this group, the activation intensity of the occipital cortex correlated significantly with MBI improvement, likely driven by the continuous input of visual cues, such as real-time feedback on robotic movement trajectories during passive mode training. This process appears to promote adaptive reorganization within the visual-spatial processing network. Meanwhile, patients in the conventional group exhibited unaffected-side dominant activation patterns consistent with prior studies. However, their MBI improvement still showed a significant correlation with activation in the affected-side M1 (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.659, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.006), further affirming the critical role of engaging the residual network on the affected side in facilitating functional recovery [\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e].\u003c/p\u003e \u003cp\u003ePhase synchronization analysis revealed that different training modes employ distinct strategies for regulating the reorganization of brain network connectivity. Following treatment, the assistive mode group demonstrated significant enhancement of intra-hemispheric frontal-motor network connectivity on the affected side, particularly by strengthening functional links between the ipsilesional PFC, DLPFC, and M1. These improvements may underpin a closed-loop control mechanism that incorporates motor planning, error monitoring, and real-time adjustments [\u003cspan additionalcitationids=\"CR53\" citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eConversely, the passive mode group showed increased connectivity within the cross-hemispheric sensory-visual pathway, such as between the contralesional SFC and the ipsilesional OC. This likely reflects a dependency on multisensory integration pathways but reveals a lack of improvement in direct motor-related connections, thereby constraining motor function recovery. Notably, the conventional group and the assistive mode group displayed similarities in cross-hemispheric motor network connectivity, such as links between the contralesional PMC and the ipsilesional M1, suggesting that both active participation training modes share certain mechanisms for cross-hemispheric information transmission. However, the assistive mode group, supported by high-precision, robotics-assisted repetitive training, appears to have further amplified these effects.\u003c/p\u003e \u003cp\u003eAt the neuroelectrophysiological level, while no statistically significant changes were observed in RMT, MEP latency, or MEP amplitude across the different groups, all groups exhibited a general trend of improvement in RMT. This finding suggests that short-term training may primarily enhance motor output function by improving synaptic efficacy plasticity, rather than inducing structural remodeling of the corticospinal tract. This conclusion aligns with the findings of Reis et al. regarding the underlying mechanisms of short-term rehabilitation interventions [\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eAt the level of clinical application and translation, this study proposes a three-tier decision-making framework based on \"injury severity\u0026ndash;functional dimension\u0026ndash;neural mechanisms.\" For patients with severely impaired baseline motor functions (FAC grade 0\u0026ndash;1), assistive training is identified as the optimal approach, demonstrating clear clinical advantages by reconstructing the \"motor intention\u0026ndash;execution\" pathway. For patients with mild to moderate functional impairments, equivalent training modes may be selected based on the allocation of medical resources. Furthermore, the study highlights dimension-specific effects of various training modes on functional recovery. Specifically, motor function (FMA) improvement is more reliant on prefrontal-motor network coactivation induced by assistive mode training, balance function (BBS) enhancement primarily depends on sensory input mechanisms bolstered by passive mode training, while progress in activities of daily living (MBI) is exclusively driven by active engagement modes [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Notably, a strong correlation was observed between near-infrared imaging features, such as enhanced prefrontal synchrony and interhemispheric connectivity in the occipital lobe, and functional prognosis. These findings provide a theoretical basis for the development of neuroimaging biomarkers and point to future applications, such as constructing machine learning models based on baseline brain activation patterns to enable the precise design of individualized rehabilitation plans.\u003c/p\u003e \u003cp\u003eThis study has several limitations. The small sample size (n\u0026thinsp;=\u0026thinsp;44) may hinder the detection of certain inter-group effects, such as the main effect of FMA between groups (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.074), highlighting the need for larger multicenter studies to validate the robustness of the findings. Additionally, the two-week follow-up period restricted observation of long-term neuroplasticity effects. Future studies should consider extending this period to 3\u0026ndash;6 months to evaluate functional maintenance effects and the dynamic processes of white matter remodeling. From a methodological perspective, the current spatial resolution of near-infrared technology limits the ability to analyze mechanisms in deeper brain regions. This necessitates integrating functional magnetic resonance imaging and diffusion tensor imaging techniques to further elucidate structure-function coupling patterns. Expanding the dimensionality of MEP parameter analysis (e.g., dynamic evaluation of motor threshold-output curves) is also recommended to detect subtle changes in pyramidal tract pathways.\u003c/p\u003e \u003cp\u003eFuture research should adopt a multidimensional approach by deeply integrating multimodal neuroimaging technologies, longitudinally tracking transition patterns between task-state and resting-state functional statuses, and incorporating neuroregulation methods such as transcranial magnetic stimulation and functional electrical stimulation. Establishing a tripartite precision rehabilitation system\u0026mdash;\"exoskeleton-brain-computer interface-neuroregulation\"\u0026mdash;could leverage synergistic mechanisms to facilitate neurocompensation and remodeling. Such advancements would offer robust, evidence-based support for personalized treatment, driving progress in rehabilitation science.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study highlights that conventional therapies, robot-assisted gait training in passive mode, and robot-assisted gait training in assistive mode can significantly enhance motor function, balance ability, and daily living skills in stroke patients within a short timeframe. The varying activation patterns and neural network connections observed across different training modes suggest that these approaches may influence recovery through distinct neural mechanisms. Notably, assistive mode training demonstrates specific advantages in improving motor function and promoting neural activity reconstruction on the impaired side, whereas passive mode training uniquely contributes to enhancing balance ability and visual compensation pathways. Future research should incorporate advanced neuroimaging techniques and extended follow-up periods to examine the long-term impact of these rehabilitation strategies on brain remodeling and clinical outcomes. Such efforts would provide a more robust theoretical foundation for developing personalized rehabilitation plans.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eRAGT, Robot-assisted gait training; DLPFC, dorsolateral prefrontal cortex; M1, primary motor cortex; S1, primary somatosensory cortex; fNIRS, functional near-infrared spectroscopy; FMA-LE, Fugl-Meyer Assessment for the Lower Extremity; BBS, Berg Balance Scale; MBI, Modified Barthel Index; FAC, Functional Ambulation Category; RMT, resting motor threshold; MEP, motor-evoked potential; PFC, prefrontal cortex; SFC, superior frontal cortex; PMC, premotor cortex; OC, occipital cortex; HbO, oxygenated hemoglobin; PCA, principal component analysis; ICA, independent component analysis; WA, Wavelet Amplitude; WPCO, wavelet phase coherence; ANOVA, analysis of variance; LSD, least significant difference, SD, Standard Deviation; IQR, interquartile range; MD, mean difference; MCID, minimal clinically important difference\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eEthics approval and consent to participate\u003c/p\u003e\n\u003cp\u003eThis study was approved by the Medical Ethics Committee of the Third Affiliated Hospital of Sun Yat-sen University ([2021]02-333-03) in accordance with the declaration of Helsinki. All participants gave written informed consent prior to their enrollments.\u003c/p\u003e\n\u003cp\u003eConsent for publication\u003c/p\u003e\n\u003cp\u003eConsent from all authors has been acquired prior to submission of this article.\u003c/p\u003e\n\u003cp\u003eAvailability of data and materials\u003c/p\u003e\n\u003cp\u003eThe data that support the findings of this study are available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003eCompeting interests\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003eFunding\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003eAuthors\u0026apos; contributions\u003c/p\u003e\n\u003cp\u003eYJ.X. performed the conceptualization, investigation, drafted and wrote the main manuscript text. MS.S., XD.M., KX.C., and H.X. performed the experimental data acquisition, statistical analysis and prepared the figures. H.X. prepared the result visualization. ZL.D. is involved in project administration, as well as supervision and funding acquisition. ZL.D., and K.L were responsible for validation, reviewing, and editing the original work. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003eAcknowledgements\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003eAuthors\u0026apos; information\u003c/p\u003e\n\u003cp\u003e\u003csup\u003e1\u003c/sup\u003eDepartment of Rehabilitation Medicine, The Third Affiliated Hospital, Sun Yat-Sen University, Guangzhou, 510630, People\u0026rsquo;s Republic of China,\u003c/p\u003e\n\u003cp\u003e\u003csup\u003e2\u003c/sup\u003eGuangzhou International Economics College, Guangzhou, 510450, People\u0026rsquo;s Republic of China\u003c/p\u003e\n\u003cp\u003e\u003csup\u003e3\u003c/sup\u003eDepartment of Biomedical Engineering, Faculty of Engineering, Hong Kong Polytechnic University, Hong Kong, 999077, SAR, People\u0026rsquo;s Republic of China\u003c/p\u003e\n\u003cp\u003e\u003csup\u003e4\u003c/sup\u003eDepartment of Hearing and Speech Science, Guangzhou Xinhua University, Guangzhou, 510520, People\u0026rsquo;s Republic of China\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eCampagnini S, et al. Effects of control strategies on gait in robot-assisted post-stroke lower limb rehabilitation: a systematic review. J Neuroeng Rehabil. 2022;19(1):52.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLin YN, et al. Hybrid robot-assisted gait training for motor function in subacute stroke: a single-blind randomized controlled trial. J Neuroeng Rehabil. 2022;19(1):99.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eElmas Bodur B, Erdoğanoğlu Y, Asena S, Sel. Effects of robotic-assisted gait training on physical capacity, and quality of life among chronic stroke patients: A randomized controlled study. J Clin Neurosci. 2024;120:129\u0026ndash;37.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeow XRG, Ng SLA, Lau Y. Overground Robotic Exoskeleton Training for Patients With Stroke on Walking-Related Outcomes: A Systematic Review and Meta-analysis of Randomized Controlled Trials. Arch Phys Med Rehabil. 2023;104(10):1698\u0026ndash;710.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang H, et al. Effect and optimal exercise prescription of robot-assisted gait training on lower extremity motor function in stroke patients: a network meta-analysis. Neurol Sci. 2025;46(3):1151\u0026ndash;67.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuo C, et al. Effectiveness of unilateral lower-limb exoskeleton robot on balance and gait recovery and neuroplasticity in patients with subacute stroke: a randomized controlled trial. J Neuroeng Rehabil. 2024;21(1):213.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang J, et al. Effect of robotic exoskeleton training on lower limb function, activity and participation in stroke patients: a systematic review and meta-analysis of randomized controlled trials. Front Neurol. 2024;15:1453781.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJiang Z, et al. Effects of body weight support training on balance and walking function in stroke patients: a systematic review and meta-analysis. Front Neurol. 2024;15:1413577.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ede Miguel-Fern\u0026aacute;ndez J, et al. Control strategies used in lower limb exoskeletons for gait rehabilitation after brain injury: a systematic review and analysis of clinical effectiveness. J Neuroeng Rehabil. 2023;20(1):23.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThimabut N, et al. Effects of the Robot-Assisted Gait Training Device Plus Physiotherapy in Improving Ambulatory Functions in Patients With Subacute Stroke With Hemiplegia: An Assessor-Blinded, Randomized Controlled Trial. Arch Phys Med Rehabil. 2022;103(5):843\u0026ndash;50.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMurata A, Wen W, Asama H. The body and objects represented in the ventral stream of the parieto-premotor network. Neurosci Res. 2016;104:4\u0026ndash;15.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKleim JA, Jones TA. Principles of Experience-Dependent Neural Plasticity: Implications for Rehabilitation After Brain Damage. J Speech Lang Hear Res. 2008;51(1):S225\u0026ndash;39.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSong KJ, et al. The effect of robot-assisted gait training on cortical activation in stroke patients: A functional near-infrared spectroscopy study. NeuroRehabilitation. 2021;49(1):65\u0026ndash;73.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePeters S, et al. Passive, yet not inactive: robotic exoskeleton walking increases cortical activation dependent on task. J Neuroeng Rehabil. 2020;17(1):107.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBonnal J et al. Relation between Cortical Activation and Effort during Robot-Mediated Walking in Healthy People: A Functional Near-Infrared Spectroscopy Neuroimaging Study (fNIRS). Sens (Basel), 2022. 22(15).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLouie DR, et al. Efficacy of an exoskeleton-based physical therapy program for non-ambulatory patients during subacute stroke rehabilitation: a randomized controlled trial. J Neuroeng Rehabil. 2021;18(1):149.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHirano S, et al. Effects of robot-assisted gait training using the Welwalk on gait independence for individuals with hemiparetic stroke: an assessor-blinded, multicenter randomized controlled trial. J Neuroeng Rehabil. 2024;21(1):76.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTam PK, et al. Overground robotic exoskeleton vs conventional therapy in inpatient stroke rehabilitation: results from a pragmatic, multicentre implementation programme. J Neuroeng Rehabil. 2025;22(1):3.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee SY, et al. Effects of robot-assisted walking training on balance, motor function, and ADL depending on severity levels in stroke patients. Technol Health Care. 2024;32(5):3293\u0026ndash;307.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchulz KF, Altman DG, Moher D. CONSORT 2010 statement: updated guidelines for reporting parallel group randomised trials. BMJ. 2010;340:c332.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLouie DR, et al. Efficacy of an exoskeleton-based physical therapy program for non-ambulatory patients during subacute stroke rehabilitation: a randomized controlled trial. J Neuroeng Rehabil. 2021;18(1):149.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaker WB, et al. Modified Beer-Lambert law for blood flow. Biomed Opt Express. 2014;5(11):4053\u0026ndash;75.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTan Q, et al. Frequency-specific functional connectivity revealed by wavelet-based coherence analysis in elderly subjects with cerebral infarction using NIRS method. Med Phys. 2015;42(9):5391\u0026ndash;403.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePandian S, Arya KN, Kumar D. Minimal clinically important difference of the lower-extremity fugl-meyer assessment in chronic-stroke. Top Stroke Rehabil. 2016;23(4):233\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu CY, et al. Constraint-induced therapy with trunk restraint for improving functional outcomes and trunk-arm control after stroke: a randomized controlled trial. Phys Ther. 2012;92(4):483\u0026ndash;92.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTaghavi A, Sharabiani P, et al. Minimal important difference of Berg Balance Scale, performance-oriented mobility assessment and dynamic gait index in chronic stroke survivors. J Stroke Cerebrovasc Dis. 2024;33(11):107930.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCramer SC. Repairing the human brain after stroke: I. Mechanisms of spontaneous recovery. Ann Neurol. 2008;63(3):272\u0026ndash;87.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKleim JA, Jones TA. Principles of experience-dependent neural plasticity: implications for rehabilitation after brain damage. J Speech Lang Hear Res. 2008;51(1):S225\u0026ndash;39.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePollock A et al. Interventions for improving upper limb function after stroke. Cochrane Database Syst Rev, 2014. 2014(11): p. Cd010820.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLanghorne P, Coupar F, Pollock A. Motor recovery after stroke: a systematic review. Lancet Neurol. 2009;8(8):741\u0026ndash;54.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWinstein CJ, et al. Guidelines for Adult Stroke Rehabilitation and Recovery: A Guideline for Healthcare Professionals From the American Heart Association/American Stroke Association. Stroke. 2016;47(6):e98\u0026ndash;169.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCalautti C, Baron JC. Functional neuroimaging studies of motor recovery after stroke in adults: a review. Stroke. 2003;34(6):1553\u0026ndash;66.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKrakauer JW. Motor learning: its relevance to stroke recovery and neurorehabilitation. Curr Opin Neurol. 2006;19(1):84\u0026ndash;90.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePascual-Leone A et al. \u003cem\u003eTHE PLASTIC HUMAN BRAIN CORTEX.\u003c/em\u003e Annual Review of Neuroscience, 2005. 28(Volume 28, 2005): pp. 377\u0026ndash;401.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiepert J, et al. Treatment-Induced Cortical Reorganization After Stroke in Humans. Stroke. 2000;31(6):1210\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWard NS, et al. Motor system activation after subcortical stroke depends on corticospinal system integrity. Brain. 2006;129(3):809\u0026ndash;19.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrefkes C, et al. Cortical connectivity after subcortical stroke assessed with functional magnetic resonance imaging. Ann Neurol. 2008;63(2):236\u0026ndash;46.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoeller K, Willmes K, Klein E. A review on functional and structural brain connectivity in numerical cognition. Front Hum Neurosci, 2015. 9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHampson M, et al. Brain Connectivity Related to Working Memory Performance. J Neurosci. 2006;26(51):13338\u0026ndash;43.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMihara M, et al. Neurofeedback Using Real-Time Near-Infrared Spectroscopy Enhances Motor Imagery Related Cortical Activation. PLoS ONE. 2012;7(3):e32234.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBadcock JC, Hugdahl K. A synthesis of evidence on inhibitory control and auditory hallucinations based on the Research Domain Criteria (RDoC) framework. Front Hum Neurosci, 2014. 8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTakeuchi N, et al. Repetitive Transcranial Magnetic Stimulation of Contralesional Primary Motor Cortex Improves Hand Function After Stroke. Stroke. 2005;36(12):2681\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMurase N, et al. Influence of interhemispheric interactions on motor function in chronic stroke. Ann Neurol. 2004;55(3):400\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchaechter JD. Motor rehabilitation and brain plasticity after hemiparetic stroke. Prog Neurobiol. 2004;73(1):61\u0026ndash;72.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKitago T, Krakauer JW. Chap. 8 \u003cem\u003e- Motor learning principles for neurorehabilitation\u003c/em\u003e. In: Barnes MP, Good DC, editors. Handbook of Clinical Neurology. Editors: Elsevier; 2013. pp. 93\u0026ndash;103.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKawai R, et al. Motor Cortex Is Required for Learning but Not for Executing a Motor Skill. Neuron. 2015;86(3):800\u0026ndash;12.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGladstone DJ, Danells CJ, Black SE. The fugl-meyer assessment of motor recovery after stroke: a critical review of its measurement properties. Neurorehabil Neural Repair. 2002;16(3):232\u0026ndash;40.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMahoney FI, Barthel DW. FUNCTIONAL EVALUATION: THE BARTHEL INDEX. Md State Med J. 1965;14:61\u0026ndash;5.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLang CE, MacDonald JR, Gnip C. Counting Repetitions: An Observational Study of Outpatient Therapy for People with Hemiparesis Post-Stroke. J Neurol Phys Ther. 2007;31(1):3\u0026ndash;10.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eContreras-Vidal JL, et al. Neural Decoding of Robot-Assisted Gait During Rehabilitation After Stroke. Am J Phys Med Rehabil. 2018;97(8):541\u0026ndash;50.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarter AR, et al. Resting interhemispheric functional magnetic resonance imaging connectivity predicts performance after stroke. Ann Neurol. 2010;67(3):365\u0026ndash;75.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSitaram R, et al. Closed-loop brain training: the science of neurofeedback. Nat Rev Neurosci. 2017;18(2):86\u0026ndash;100.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRidderinkhof KR, et al. The Role of the Medial Frontal Cortex in Cognitive Control. Science. 2004;306(5695):443\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHamada M, et al. The Role of Interneuron Networks in Driving Human Motor Cortical Plasticity. Cereb Cortex. 2012;23(7):1593\u0026ndash;605.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eReis J et al. \u003cem\u003eNoninvasive cortical stimulation enhances motor skill acquisition over multiple days through an effect on consolidation.\u003c/em\u003e Proceedings of the National Academy of Sciences, 2009. 106(5): pp. 1590\u0026ndash;1595.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Robot-Assisted Gait Training, Stroke, Functional Near-Infrared Spectroscopy, Motor Function, Neuroplasticity, Cortical Activation, Functional Connectivity","lastPublishedDoi":"10.21203/rs.3.rs-6535268/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6535268/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eRobot-assisted gait training (RAGT) has gained recognition as a promising therapeutic approach, offering high-intensity and repetitive training. Despite its potential, the clinical effectiveness and ideal training protocols continue to be subjects of debate. This study seeks to compare the impacts of various RAGT modes on lower limb motor function recovery in stroke patients while exploring the corresponding neural mechanisms.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eA double-blind, randomized controlled trial was conducted on patients aged 18 to 80 who had experienced their first unilateral stroke accompanied by walking impairments. Participants were randomly assigned to one of three groups: (1) assistive mode training combined with conventional therapies, (2) passive mode training combined with conventional therapies, or (3) a control group receiving only traditional rehabilitation. Outcomes were evaluated using the Fugl-Meyer Assessment for Lower Extremity (FMA-LE), Berg Balance Scale (BBS), Modified Barthel Index (MBI), the Functional Ambulatory Category (FAC), and functional near-infrared spectroscopy. Statistical analyses were performed using repeated measures ANOVA and non-parametric tests where appropriate, with statistical significance set at p\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eAmong the 48 patients recruited, significant time effects were observed across all groups in FMA-LE scores (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Notable improvements were detected in the conventional group (MD\u0026thinsp;=\u0026thinsp;2.688, p\u0026thinsp;=\u0026thinsp;0.005) and the passive group (MD\u0026thinsp;=\u0026thinsp;3.667, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), with the assistive mode also demonstrating a significant effect (MD\u0026thinsp;=\u0026thinsp;1.789, p\u0026thinsp;=\u0026thinsp;0.039). BBS scores improved across all groups; however, no significant differences were noted between the groups (p\u0026thinsp;=\u0026thinsp;0.106). Similarly, MBI scores showed a significant time effect (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), without notable group differences (p\u0026thinsp;=\u0026thinsp;0.286). Crucially, the assistive mode training group exhibited significant differences in brain activation during tasks in specific regions compared to the control group, alongside notable interhemispheric connectivity differences.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eAll training modalities effectively enhanced motor function, balance, and daily living skills in stroke patients. Nevertheless, unique activation patterns and neural pathways suggest distinct underlying mechanisms for each training approach.\u003c/p\u003e\u003ch2\u003eTrial Registration:\u003c/h2\u003e \u003cp\u003eThe study was registered with the China Clinical Trial Registration Center under the trial registration number ChiCTR2100054527.\u003c/p\u003e","manuscriptTitle":"Comparative Effectiveness of Passive vs. Assistive Robotic Gait Training on Functional Recovery and Neuroplasticity Post-Stroke: A Randomized Controlled Trial","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-09 01:28:29","doi":"10.21203/rs.3.rs-6535268/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"cc0b913d-b4b8-4ac5-9441-62fca984c19f","owner":[],"postedDate":"May 9th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-12-15T07:24:33+00:00","versionOfRecord":[],"versionCreatedAt":"2025-05-09 01:28:29","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6535268","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6535268","identity":"rs-6535268","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}