This review proposes a cohesive closed-loop neuromorphic strategy that integrates multiple subdomains of brain-computer interface (BCI) research to address the clinical challenges hindering the translation of implantable systems. Despite advances in invasive BCI technology, conventional CMOS-based approaches remain limited by high power consumption, significant tissue rejection, and signal attenuation, which collectively compromise long-term implantation efficacy. Our approach unites key areas, including neuromorphic computing, spiking neuron devices, memristive crossbar arrays, and biocompatible interfaces, into a unified framework designed as a technical roadmap to overcome these barriers. By leveraging the inherent advantages of neuromorphic systems, such as high energy efficiency, event-driven operation, and bio-inspired architectures, this strategy aims to guide the construction of fully closed-loop BCIs capable of real-time neural signal processing and adaptive intervention. Furthermore, we explore the potential of such closed-loop systems to drive precise neuromodulation and functional restoration in neurosurgical diseases, offering a forward‑looking, promising perspective on the treatment of epilepsy, Parkinson’s disease, and spinal cord injury. Finally, we discuss future directions, emphasizing the integration of emerging technologies and interdisciplinary collaboration to advance clinical translation. This review highlights the promise of neuromorphic closed-loop strategies in bridging neuromorphic engineering with clinical neurosurgery, pointing toward next-generation intelligent, low-power, and clinically viable BCI systems.
Assessing skeletal muscle fatigue is essential for diagnosing neuromuscular impairment, but conventional methods rely on maximal or tetanic contractions that can be impractical or uncomfortable in clinical populations. Surface mechanomyography (MMG) provides a non-invasive alternative; however, current MMG-based fatigue protocols require time-consuming manual peak-to-peak analysis. This study evaluated an automated algorithm for extracting MMG-derived fatigue metrics from electrically evoked muscle twitches. Eighteen healthy adults completed a standardized fatigue protocol on the wrist extensors and ankle dorsiflexors using electrical stimulation at 2, 4, and 6 Hz over 9 min. A triaxial accelerometer captured MMG signals from > 2,100 contractions per muscle (approximately 4,320 per participant across the two muscles). A custom algorithm automatically extracted peak-to-peak values and computed the endurance index; the mean contraction amplitude was derived from peak and trough points identified manually by the research team. Repeated-measures ANOVAs assessed differences across stimulation frequencies and muscle groups. Endurance index declined significantly across the fixed ascending 2-, 4-, and 6-Hz stimulation sequence (p < 0.001) and was lower in ankle dorsiflexors than wrist extensors (p = 0.010), averaging approximately 6
Chronic ankle instability (CAI) is a common musculoskeletal condition accompanying sensorimotor dysfunction. Recent research indicates that people with unilateral CAI not only present impairments on the injured side, but also exhibit deficits on the contralateral side. Studies have observed abnormal muscle activation and co-contractions in people with CAI, suggesting potential muscle coordination disorders. Considering the observed bilateral dysfunction, this study aimed to investigate characteristics of bilateral motor control and muscle coordination via intermuscular coherence (IMC) in people with unilateral CAI. Thirty participants with unilateral CAI and 30 healthy controls were recruited. Surface electromyography of muscle pairs in bilateral limbs was measured during single-leg stance and landing: rectus femoris-biceps femoris, tibialis anterior (TA)-peroneus longus (PL), TA-gastrocnemius medial, and TA-soleus (SOL). The sway velocity of center of pressure (COPV) was calculated for single-leg stance recorded by force plate, and joint coordination pattern and variability were analyzed for hip sagittal-ankle sagittal, hip sagittal-ankle frontal, knee sagittal-ankle sagittal, knee sagittal-ankle frontal (KSAF) and knee frontal-ankle frontal coupling pairs before and after landing. CAI group displayed postural instability (anteroposterior: mean difference (MD) = 3.87 mm/s, mediolateral: MD = 5.52 mm/s), higher KSAF variability (MD = 1.72˚), higher TA-PL β (MD = 0.035) and TA-SOL β (MD = 0.024) IMC bilaterally. An increased number of ankle sprains were correlated with higher IMC during post-landing bilaterally (injured: ꞵ = 0.398, adjusted R2 = 0.129, uninjured: ꞵ = 0.387, adjusted R2 = 0.120). Associations between IMC and COPV and coordination variability for both groups were observed (p ≤ 0.044). There were also correlations of IMC between two sides in unilateral CAI group (p ≤ 0.041), especially for the TA-SOL muscle pair. People with unilateral CAI displayed higher variability and IMC during landing bilaterally, suggesting abnormal motor regulation and additional central involvement. The number of ankle sprains and function on the injured side may modulate the function on the uninjured side. Rehabilitation or training for people with unilateral CAI may need to transition from injured side to bilateral sides or employ methods that could modulate bilateral functions.
This study introduces an optimal control framework for functional electrical stimulation (FES) cycling that produces personalized stimulation patterns directly transferable from simulation to experimental cycling, without the need for empirical tuning. We use a physiological muscle model that accounts for the effects of electrical stimulation on muscle contraction dynamics, whose parameters are identified using experimental data obtained from a dedicated protocol conducted with a participant with spinal cord injury (SCI). We demonstrate the feasibility of obtaining a stimulation pattern via trajectory optimization to track a 3.6 rad/s (35 rpm) cadence and a 25 W crank power output by optimizing muscle stimulation pulse durations. The simulations are conducted using two different models—one comprising six muscles and the other four—to explore different experimental conditions. Finally, we assess whether the resulting stimulation patterns enable the participant to pedal a recumbent tricycle without further modification. The generated patterns successfully produce continuous, coordinated pedaling. The average power output was 18 W with a mean angular velocity of 3.93 ± 0.80 rad/s for the six-muscle configuration, and 19 W with a mean angular velocity of 3.90 ± 1.13 rad/s for the four-muscle configuration. This study introduces a physiological model-based control approach for FES cycling, bridging musculoskeletal simulation and experimental validation, and underscores the potential of predictive simulations and optimal control to personalize FES-assisted protocols.
Temporal interference stimulation (TIS) is a promising non-invasive technique for reaching deep brain regions that are difficult to modulate with conventional transcranial electrical stimulation. A critical question is whether TIS can reliably induce neural modulation at its envelope frequency in humans. As an initial step toward applications targeting deeper structures, we investigated whether TIS can modulate cortical oscillations at the envelope frequency within the primary somatosensory cortex (S1). We also examined how electric fields in off-target regions contribute to interindividual variability in stimulation efficacy using individualized simulations based on each participant’s magnetic resonance imaging (MRI). Forty-nine healthy participants were enrolled (24 TIS, 25 active sham without envelope modulation). TIS was applied over the left S1 hand area using a stimulation protocol designed to generate a 10 Hz envelope frequency. Brain activity was recorded with magnetoencephalography (MEG) before and after stimulation. Electric field simulations were conducted using individualized head models reconstructed from each participant’s structural MRI. The primary mixed two-way ANOVA revealed no significant group-by-time interaction for alpha-band power, with a sensitivity analysis indicating that the study was slightly underpowered to detect this interaction effect. Subsequent exploratory within-group analyses showed a post-stimulation increase in alpha band power around 10 Hz in the targeted S1 in the TIS group (Cohen’s d = 0.570), whereas no significant change was observed in the active sham group. No clear changes were observed in the beta or gamma bands, and the increase was descriptively largest in the alpha band, although a direct statistical comparison across frequency bands did not reach significance. Furthermore, exploratory analyses based on individualized electric field simulations suggested that off-target electric fields may attenuate the relationship between local field strength in S1 and changes in alpha-band oscillations. These findings provide preliminary evidence suggesting that TIS may modulate cortical oscillations in humans in accordance with the envelope frequency, although the evidence remains limited. Furthermore, its effectiveness may be influenced by off-target electric fields.
The prevention of pressure injury (PI) remains challenging due to limited clinical evidence, while existing biomechanical studies, as complementary approaches, are largely confined to static conditions, simplified models, and insufficient representation of dynamic mattress behavior. Therefore, this study developed a multi-scale biomechanical framework to evaluate buttocks–support surface interactions under both static and dynamic loading conditions and to inform surface selection and design. A subject-specific buttocks model reconstructed from MRI/CT was combined with three support surfaces modeled in CAD—standard foam mattress (SFM), strip air-filled mattress (ST-AFM), and spherical air-filled mattress (SP-AFM)—to generate three distinct biomechanical models. Soft tissues and SFM were modeled using hyperelastic Ogden formulations, while air-chamber membranes were represented by calibrated Mooney–Rivlin models with internal ideal gas. Finite element simulations were conducted under static and alternating-pressure conditions (24/12 kPa), evaluating stiffness, reaction force, interface stress, and tissue strain. Under static loading, SFM exhibited the highest stiffness (7.69 N/mm), whereas SP-AFM showed lower stiffness (5.35 N/mm) and a lower proportion of simulated high-stress regions (> 16 kPa) by up to 0.53
Cerebral palsy (CP) is a nonprogressive brain disorder that causes permanent movement and posture impairments, often resulting in gait dysfunction. Improving walking ability is a key therapeutic goal, as it enhances daily functioning and social participation. Gait quality, walking speed, and endurance are commonly assessed using the Gait Deviation Index (GDI), gait velocity (v), and the six-minute walk test (6MWT). Although these measures are widely used, it remains unclear how strongly they are related to one another and whether their changes over time follow similar patterns. Importantly, each feature is assessed with different procedures and practical constraints, which may limit joint interpretation in routine clinical practice. Therefore, this study aimed to quantify the relationships between GDI, gait velocity, and distance measured during the 6MWT (6MWT distance) in ambulatory individuals with CP and to determine whether their changes follow a similar pattern over time. In total, 360 ambulatory participants with spastic CP (Gross Motor Function Classification System (GMFCS) I–III) were assessed twice over a 24-month period (T1: first assessment; T2: 24 months later; paired T1–T2 data were available for N=211). Gait quality (GDI), walking speed (velocity), and endurance (6MWT distance) were measured using instrumented gait analysis and standardized protocols. Correlations between these measures were analysed using Pearson’s or Spearman’s coefficients, with significance set at p < 0.05. At baseline (T1), in the entire group, the GDI was moderately correlated with gait velocity (p < 0.01, r = 0.59) and strongly correlated with 6MWT distance (p < 0.01, r = 0.61); gait velocity and 6MWT distance were also strongly correlated (p < 0.01, r = 0.61). In subgroup analyses, significant baseline correlations of varying strength were observed across all the subgroups. For changes from T1 to T2, correlations in the entire group were very weak for changes in GDI vs. changes in gait velocity (p = 0.03, r = 0.15) and for changes in GDI vs. changes in 6MWT distance (p = 0.02, r = 0.16); changes in gait velocity had a stronger correlation with changes in the 6MWT, but this correlation was still weak (p < 0.01, r = 0.26). Significant correlations between changes in the mentioned parameters occurred only in selected subgroups, again with varying strength. Gait quality, walking speed, and gait endurance are correlated in individuals with CP, and the strength of these relationships depends on the level of ambulatory function. Although gait quality changed over time, its change showed only very weak associations with changes in walking speed and gait endurance; in independently ambulant participants (GMFCS I–II), changes in the GDI were not associated with changes in velocity or walking distance, whereas in participants with GMFCS III, improvements in gait quality aligned with longer walking distance but not with higher velocity. Overall, these findings support a pragmatic, individualized approach to clinical assessment, guided by patients’ ambulatory function level and the limitations of the GDI for interpreting longitudinal change. Institutional Review Board permission no. 767/24.
Together with the aging of the world population, stroke incidence is expected to increase in the coming years. Stroke is frequently associated with sensorimotor impairments of the upper limb, including proprioceptive deficits, that substantially impair motor control and quality of life. Understanding how physiological aging and acute stroke affect proprioceptive processing is therefore crucial for developing targeted rehabilitation strategies. Vibration-induced illusion of movement was administered to 28 young, 30 older, and 23 individuals with acute stroke to the distal tendon of the triceps brachii. In healthy participants, vibration was applied to both arms, whereas in participants with acute stroke, it was applied to the paretic left arm. Vibration-induced illusion of movement was delivered either with or without visual feedback of the vibrated arm. Each condition consisted of six 15-s vibrations at 80 Hz. The illusion was quantified using the Standardized Kinesthetic Illusion Procedure scale, while cerebral hemodynamic activity within frontoparietal and sensorimotor areas was assessed using fNIRS through changes in (de)oxyhemoglobin concentrations. Subsequently, activation profiles were established, and group cerebral hemodynamics and functional neural efficiency were compared. Older adults exhibited more extensive bilateral activation in the frontoparietal and sensorimotor cortices compared with young adults, suggesting age-related changes in neural recruitment during proprioceptive processing. In contrast, participants with acute stroke showed no significant task-related cortical activation and displayed an overall reduction in cerebral activity compared with both healthy groups. Despite this reduced activation, neural efficiency in participants with acute stroke was comparable to that observed in older participants. Together, these findings suggest that preserved compensatory mechanisms may still be engaged early after stroke, and suggest potential implications of vibration-induced illusion of movement to limit age-related proprioceptive decline and to support early rehabilitation after stroke. Clinical Trial registration: NCT06218563—2024-01-12.
The role of the right hemisphere in aphasia recovery remains a central debate in neurorehabilitation. While traditional paradigms often focus on a binary “excitation vs. inhibition” logic, the dynamic nature of the subacute recovery window suggests that a more nuanced, state-dependent approach is required. In this study, we investigated the efficacy and neural mechanisms of high-frequency (excitatory) repetitive transcranial magnetic stimulation (HF-rTMS) targeting the right inferior frontal gyrus in subacute post-stroke aphasia. Participants received 18 sessions of active or sham HF-rTMS over three weeks, integrated with conventional speech-language therapy. Neurophysiological evaluation was performed using functional near-infrared spectroscopy (fNIRS) during a Verbal Fluency Task (VFT) to measure task-related hemodynamic responses (ΔHbO, ΔDHb and ΔtHb) and interhemispheric dynamics at baseline and post-intervention. Active HF-rTMS yielded significantly greater gains in the Aphasia Quotient (AQ) compared to sham stimulation. Neurally, the active group exhibited a specific modulation of the right inferior frontal gyrus; however, rather than a uniform increase in excitability, we identified a “differential adaptive spectrum” consisting of both hemodynamic “increasers” and " decreasers.” These heterogeneous responses correlated with distinct behavioral recovery trajectories, suggesting that the same excitatory stimulus can facilitate different compensatory pathways depending on the individual’s baseline network state. Conclusions: High-frequency stimulation of the right inferior frontal gyrus (IFG) may support language recovery in subacute post-stroke aphasia, but the accompanying hemodynamic responses appear to differ across individuals. These findings indicate that the functional role of the right hemisphere cannot be reduced to a uniformly maladaptive or compensatory process. Accordingly, inhibitory and excitatory stimulation of the right hemisphere may not be equally appropriate for all patients. Future studies should determine whether baseline clinical and neurophysiological profiles can be used to guide more individualized selection of stimulation polarity and target.
Persistent wrist-hand motor impairment is common after stroke. Whether combining repetitive transcranial magnetic stimulation (rTMS) with task-oriented soft robotic glove (SRG) training provides additional clinical benefit remains uncertain. This study evaluated the effects of rTMS combined with SRG training on upper-limb motor recovery after subacute stroke and explored associated changes in cortical connectivity and corticomuscular information transfer. In this single-center, assessor-blinded, three-arm randomized controlled trial, 87 participants with subacute stroke were randomly assigned (1:1:1) to rTMS + SRG group, sham rTMS + SRG group, or rTMS combined with time-matched conventional occupational therapy, in addition to standardized background rehabilitation, for 2 weeks, and followed for 1 month. The primary outcome was the Fugl-Meyer Assessment for the Upper Extremity (FMA-UE). Secondary outcomes included FMA upper-arm and wrist-hand subscores, Action Research Arm Test, and Modified Barthel Index. Resting-state functional near-infrared spectroscopy (fNIRS) assessed functional connectivity. Task-based fNIRS and surface electromyography were used to calculate transfer entropy. A total of 79 participants completed all interventions and assessments. Baseline demographic and clinical characteristics were comparable among the three groups. Significant group-by-time interactions were observed for the primary and secondary clinical outcomes (P˂0.05). FMA-UE improvement was greater in the rTMS + SRG group than in the sham rTMS + SRG and rTMS + OT groups (mean improvement, 7.07, 4.77, and 4.65 points, respectively; Dunn-adjusted P = 0.0027 and 0.0005 for pairwise comparisons with the rTMS + SRG group). This advantage remained evident at 1-month follow-up. Improvements were most pronounced in the FMA wrist-hand subscore, with greater gains in the rTMS + SRG group than in both comparator groups during the intervention period (Dunn-adjusted P = 0.0067 and < 0.001). Resting-state fNIRS revealed greater increases in motor-related functional connectivity in the rTMS + SRG group. Transfer entropy also showed greater increases in the rTMS + SRG group than in both comparator groups (Dunn-adjusted P = 0.026 and 0.048). rTMS combined with task-oriented SRG training was associated with greater improvement in upper-limb motor impairment after subacute stroke, particularly in distal wrist-hand impairment. Changes in cortical connectivity and task-related corticomuscular information transfer provide exploratory neurophysiological insights, but their mechanistic relevance to clinical recovery requires further investigation. Trial registration Chinese Clinical Trial Registry, ChiCTR2400090007.
Physical exercise is increasingly recognized as a supportive intervention for people with Parkinson’s disease (PwPD), yet the biological mechanisms underlying its effects remain unclear. This study investigated whether exercise-based rehabilitation modulates circulating biomarkers associated with neuroaxonal integrity, inflammation, and muscle–brain communication. In this randomized controlled substudy, 40 PwPD were assigned to a multidisciplinary intensive outpatient rehabilitation program (experimental group, EXP; n = 20) or a home-based self-managed intervention (control group, CTRL; n = 20). Assessments were performed at baseline (T0), post-treatment (T1), and 3-month follow-up (T2). Clinical outcomes included motor function (UPDRS-III), cognition (MoCA), and functional status (ADL, IADL). Plasma levels of neurofilament light chain (NFL), irisin, interleukin-6 (IL-6), and soluble IL-6 receptor (sIL-6R) were measured. No significant changes over time or between groups were observed in motor or cognitive outcomes. In contrast, significant biomarker modulation was detected. Irisin levels increased in the EXP group at T1 and T2, while decreasing in the CTRL group (p < 0.01). NFL levels significantly decreased in the EXP group at both time points, whereas an increase was observed at follow-up in the CTRL group. IL-6 remained stable in both groups. sIL-6R levels decreased over time, with a significant group-by-time interaction at T1 (p = 0.038). No significant correlations emerged between clinical outcomes and biomarkers. Intensive multidisciplinary rehabilitation was associated with significant modulation of circulating biomarkers, despite the absence of clinical changes. Decreased NFL and increased irisin suggest potential neuroprotective and exercise-responsive effects. These findings support the use of plasma biomarkers as sensitive indicators of neurobiological adaptations to rehabilitation in PwPD. Trial registration The original randomized controlled trial (RCT) was registered at ClinicalTrials.gov (Study ID: NCT05452655)
Neuromuscular fatigue increases variability in motor-unit (MU) discharge and force output, yet how it reorganizes population-level MU discharge dynamics remains poorly understood. Conventional analytical approaches primarily quantify discharge magnitude, variability, and common synaptic input, providing limited information regarding the organization of discharge-state space. Hence, we applied a time-resolved energy landscape framework that preserves a joint representation of the mean and variability of pooled MU discharge to characterize fatigue-related reorganization of population-level MU discharge-state space. This approach was motivated by the premise that force steadiness depends on the joint organization of neural drive level and neural drive variability. Force output and decomposed surface electromyographic signals were recorded during submaximal isometric wrist extension before and after an ischemia-assisted fatigue protocol in forty healthy adults. Thirty-five participants exhibiting post-fatigue reductions in maximal voluntary contraction force were included in subsequent analyses. Pooled MU discharge activity was represented by the instantaneous mean firing rate and analyzed using a sliding-window approach. Energy landscapes were constructed in a two-dimensional discharge-state space defined by the mean and variability of pooled MU discharge activity, enabling quantification of landscape structure (basin number, basin area, and basin depth) and state-space occupancy (centroid location and centroid dispersion). Fatigue significantly increased force fluctuation magnitude (P < 0.001) and altered the organization of MU discharge-state space. Specifically, fatigue increased the number (P = 0.017) and area (P = 0.012) of attractor basins, indicating a broader distribution of preferred discharge configurations. Fatigue also induced significant shifts in centroid location (P ≤ 0.008) and increased centroid dispersion (P = 0.004), reflecting broader occupancy of discharge states. Furthermore, fatigue-related increases in force fluctuation magnitude were positively associated with basin area expansion (r = 0.368, P = 0.030) and centroid dispersion (r = 0.468, P = 0.005). Ischemia-assisted fatigue reorganized the state-space architecture of pooled MU discharge activity, resulting in broader discharge-state occupancy and reduced force steadiness. Energy landscape analysis provides a complementary time-resolved framework that preserves the joint evolution of the mean and variability of pooled MU discharge, offering a novel state-space perspective on fatigue-related decline in force stability.
Phantom limb pain (PLP) is a debilitating condition associated with maladaptive neuroplastic changes following limb amputation. The prevailing neuromodulatory approach targets the deafferented contralateral motor cortex, yet interhemispheric imbalance is increasingly recognized as a central pathophysiological feature of PLP. We investigated whether inhibitory repetitive transcranial magnetic stimulation (rTMS) applied to the ipsilateral (unaffected) hemisphere could reduce PLP and induce measurable normalization of interhemispheric cortical asymmetry. Two individuals with traumatic right upper-limb amputation and severe PLP underwent a longitudinal study comprising a four-week baseline phase, an inhibitory 1 Hz rTMS intervention targeting the ipsilateral motor cortex, and a five-week follow-up phase. Serial navigated transcranial magnetic stimulation (nTMS) motor mapping of both hemispheres was performed before and after the intervention. A principal component analysis (PCA)-based spatial projection and Gaussian kernel regression pipeline was developed to quantify motor map area, center of gravity (CoG), mean motor evoked potential (MEP) amplitude, and signal amount bilaterally. Both participants demonstrated clinically meaningful PLP reductions (66
Spinal cord injury (SCI) profoundly disrupts motor output while also reshaping central motor systems. Motor imagery (MI) provides a valuable window into retained motor-related neural processes because it enables investigation of covert motor state generation in the absence of overt movement. However, evidence on MI after SCI remains fragmented across neurophysiological and neuroimaging modalities. The objective of this review was to systematically synthesize evidence on the neural correlates of MI after SCI across electroencephalography (EEG), functional magnetic resonance imaging (fMRI), magnetoencephalography (MEG), and intracortical recordings. A systematic search of MEDLINE, Web of Science, Scopus, Embase, and EBSCOhost was conducted from database inception to March 2026. Original studies were included if they involved individuals with SCI, used an explicit MI task, and reported MI-related neural outcomes derived from EEG, fMRI, MEG, or intracortical recordings. Because of substantial heterogeneity in injury profiles, task paradigms, modalities, and neural outcome measures, a narrative synthesis was performed. Thirty-four studies met the inclusion criteria, including 15 EEG, 10 fMRI, 4 MEG, and 5 intracortical recording studies. Across modalities, MI-related neural activity remained detectable after SCI, with evidence spanning oscillatory responses, regional recruitment, network organization, and fine-grained neural representations. Many studies also described altered recruitment patterns, modified connectivity, atypical task-related modulation, or changes in representational structure. Findings varied according to injury phenotype, chronicity, task design, and analytical framework. The four modalities provided complementary evidence on post-SCI motor imagery across temporal, spatial, network, and neuronal scales. Available evidence indicates that MI-related neural activity is generally detectable after SCI, although its organization is often altered. A multimodal perspective is therefore necessary for understanding both the persistence and the reshaping of motor representations after injury. These findings support continued investigation of MI in neurorehabilitation and neurotechnology, while also highlighting the need for approaches that account for post-injury heterogeneity.
This study compared changes in physical function following tilt-table robot-assisted training between stroke patients with and without sarcopenia. The aim was to investigate whether sarcopenia influences functional recovery during rehabilitation. A total of 74 stroke patients who received tilt-table robot-assisted training at Gwangju G Rehabilitation Hospital between July 2022 and June 2024 were included. Participants were classified into sarcopenia (SRG, n = 45) and non-sarcopenia (NSRG, n = 29) groups according to the 2019 Asian Working Group for Sarcopenia criteria. All participants underwent a 4-week tilt-table robot-assisted training program combined with the hospital's standard multidisciplinary convalescent rehabilitation program. Functional outcomes, including muscle strength (MMT), balance (BBS), gait (FAC), and activities of daily living (K-MBI), were assessed before and after the intervention. Within-group and between-group comparisons were performed using the Wilcoxon signed-rank test and Mann–Whitney U test, respectively. Multivariable linear regression analyses were performed to identify factors independently associated with post-intervention functional outcomes. Both groups showed significant improvements in muscle strength, balance, gait, and activities of daily living following the intervention (all p < .05). Participants without sarcopenia demonstrated significantly greater improvements in balance and activities of daily living than those with sarcopenia (both p < .05), whereas no between-group differences were observed for muscle strength or gait. Multivariable regression analysis demonstrated that sarcopenia remained independently associated with lower post-intervention K-MBI scores after adjustment for age, sex, baseline cognition, and baseline functional status. Both groups demonstrated significant functional improvements during the intervention period. Although participants with sarcopenia showed relatively smaller improvements in balance and activities of daily living, they also achieved meaningful functional gains. These findings suggest that individualized rehabilitation strategies may be particularly important for stroke patients with sarcopenia. Further controlled prospective studies are needed to clarify the independent effects of tilt-table robot-assisted training. Trial registration: This trial was approved by the Public Institutional Review Board of the Ministry of Health and Welfare (P01-202411-01-011) and registered in the Clinical Research Information Service of Korea (KCT0010038).
Upper limb (UL) movements are highly diverse, as are movement deficits following stroke. The relationship that stroke pathology has with UL impairments and their recovery has been studied with an increasingly complex array of neuroimaging and neurophysiology tools. This systematic review summarizes evidence on brain-related measures and UL movement in subjects with stroke. Bibliographic databases were systematically searched for combinations of ‘brain function’ and ‘kinematics/ kinetics/ muscle activity’ and ‘stroke’ and ‘upper limb’. Studies were included if both brain and UL movement measures were assessed in at least 10 subjects. Each study was evaluated for risk of bias (adapted QUADAS-2 tool). Information on assessments for brain and UL measures, their associations, and experimental tasks was extracted, summarized, and synthesized. This study was registered on PROSPERO (CRD42025637373). A total of 220 studies were included that reported 288 experimental tasks across 4'954 subjects with stroke (mean 59.7 years). Transcranial magnetic stimulation, functional magnetic resonance imaging (fMRI), and electroencephalography, commonly combined with grip force sensors and motion sensor- or camera-based systems, were the most frequently used methodologies. Overall, 92 brain-related and 201 UL metrics were assessed. Risk of bias was low-to-moderate. Structural brain measures revealed associations between corticospinal tract integrity and greater grip strength, as well as kinematic differences between left- and right-hemisphere stroke locations. For functional brain measures, increased activation of ipsilesional motor areas during fMRI was associated with improved smoothness. In addition, studies reported reduced activation intensity and excitability after stroke, alongside task-related recruitment of additional brain areas. This review summarizes a large and heterogeneous body of research reflecting substantial methodological and technological advances in the assessment of brain and UL movement measures after stroke. The current findings support the relevance and utility of combining brain-based and UL metrics to identify meaningful associations between neurological injury, brain function, and behavioral outcomes after stroke and thereby provide an improved understanding of movement after stroke.
Robotic assistance can improve the learning of complex motor skills and support rehabilitation. However, most active robotic devices developed to date provide assistance based on movement trajectories (kinematics) rather than the underlying limb dynamics (kinetics) that generate movement. To our knowledge, no study has directly compared, within a single task, how kinematics and kinetics should be integrated to enhance motor learning. The present study investigated how a complex torque–motion profile involving the right arm can be learned by manipulating torque error margins within an admittance-controlled robotic device. The participants practiced under four acquisition conditions—Constant 2 N·m, Constant 12 N·m, increasing margin (Margin+), and decreasing margin (Margin−)—in which the width of the error margin on the required torque was manipulated to induce different levels of kinetic constraints on practice. On Day 1, all the participants first completed a pretest without visual feedback, followed by 6 blocks of practice. During this acquisition phase, group-specific error margins were applied, and participants received both concurrent and terminal visual feedback of their produced torque. On Day 2, a posttest without visual feedback was performed to assess retention (identical as pretest). The results revealed that the variable-margin groups (Margin + and Margin−) led to greater improvements in kinetic and kinematic profile tracking and more stable retention than the constant-margin groups (Constant 2 N·m and Constant 12 N·m). These findings suggest that variable-margin practice enhances learning by fostering increased exploration of the task space while emphasizing both kinetic and kinematic aspects of task performance. Whereas constant margins appear to stabilize a single movement solution, variable constraints may promote a more generalizable motor control strategy, possibly by refining the predictive mapping between motor commands and their sensory consequences that contribute to updating internal models. Such insights could optimize the design of robotic rehabilitation protocols aimed at fostering durable motor recovery and relearning.
Augmented sensory feedback is expected to improve postural balance through modulation of sensory reweighting dynamics. Electrical muscle stimulation (EMS) directly assists joint movement via involuntary contractions, potentially influencing both sensory and motor systems in postural control. However, it remains unclear whether EMS-based balance training specifically facilitates adaptive changes in sensory weighting, and if so, how it modifies postural control strategies. In this study, we therefore examined the effects of balance-board training with EMS applied to the ankle muscles on sensory reweighting dynamics and postural strategies in healthy adults. Forty healthy participants performed a static standing (ST) task under four surface–vision conditions and a balance-board (BB) task requiring horizontal stabilization of the board, before and after a five-minute balance-board training session. During training, the EMS group (n = 20) received stimulation of the ankle dorsiflexor or plantarflexor muscles depending on board tilt, while the control group (n = 20) practiced without EMS. Postural outcomes were assessed using center of pressure (COP; ST task only), body segment kinematics using motion capture, and electromyography of the ankle and hip muscles. Pre–post and between-group comparisons were conducted After training, the EMS group showed increased power spectral density of the low-frequency band in the ST task, indicating increased reliance on vision. They also exhibited greater use of hip movements with shoulder stabilization, reduced coupling between ankle joint angle and COP displacements, and decreased hip muscle co-contraction compared to the control group, suggesting that EMS induced a shift toward a more coordinated hip-centered postural strategy. EMS-based balance training was associated with the induction of visual reweighting dynamics and adaptive postural strategies involving multiple joints. Specifically, EMS-induced involuntary ankle assistance elicited hip-centered adjustments in postural control, contributing to upper-trunk stabilization and increased reliance on visual inputs for postural balance regulation. These findings provide a foundational knowledge for training methods that support the learning of postural control using visual information. For establishing novel balance training tailored to individual sensory characteristics, future research should explore other augmented feedback techniques that can induce diverse forms of sensory reweighting.
Restoring dexterous hand movements in individuals with cervical spinal cord injury requires continuous control of multiple biomechanically linked degrees-of-freedom (DOF). Brain-machine interface (BMI) controlled functional electrical stimulation (FES) is a promising method to restore hand movements. We explored the suitability of restoring movement to two biomechanically linked DOF simultaneously, finger flexion and wrist flexion, using previously introduced FES methods, namely stimulation targeting individual DOFs independently. We then demonstrate a finger and wrist movement BMI that, when combined with the FES system, could restore simultaneous control of wrist and finger flexion. Two monkeys were implanted with intramuscular electrodes in muscles of the hand. Stimulation on these electrodes was used to evoke finger and wrist movements in both monkeys. Additionally, one of these monkeys was implanted with intracortical electrode arrays in the primary motor cortex and used a BMI to control wrist and finger flexion in a virtual hand without FES. First we demonstrate a monkey using an intracortical BMI to control the wrist and finger flexion in a virtual hand, both before and after the hand is temporarily paralyzed, achieving success rates and acquisition times equivalent to able-bodied control with BMI control after temporary paralysis in two sessions. We then show that intramuscular FES with two predetermined stimulation patterns, one targeting finger flexion and one targeting wrist flexion, can move the monkeys’ fingers and wrist across a functional range of motion. Due to the biomechanical coupling of the wrist and fingers, stimulation targeting finger or wrist flexion individually ultimately evokes movements in both DOF. However, we demonstrate that a simple control strategy treating each DOF independently was able to control both DOF simultaneously in a closed-loop task, achieving greater than 80
Abstract Background Hypokinetic gait increases morbidity and mortality of persons with Parkinson’s disease and constitutes a major target symptom for therapy. Feedback on gait quality, e.g. by a physiotherapist, improves stride length and reduces shuffling. Wearable devices measuring gait parameters and providing feedback could improve parkinsonian gait during everyday life when no physiotherapist is available. Here, we report on the efficacy of a new device providing discreet vibratory feedback upon decreased stride length and increased shuffling. Methods 33 persons with Parkinson’s disease, in two cohorts, received automatic vibratory feedback administered via a newly developed sensor-equipped insole when their stride length and heel strike angle decreased while walking a 730 m walking course. Gait of 14 persons with Parkinson’s disease in the first cohort was investigated in OFF and ON medication state to allow for a comparison of the feedback effect with the effect of medication on stride length, heel strike angle and gait variability. In both cohorts, the effect of feedback on gait parameters was compared against a control walk without stimulation and was subjectively evaluated by the study participants. In the second cohort of 19 participants, the effect of feedback was additionally compared with vibration at random time points to investigate the efficacy of context-adequate feedback. Results Both stride length and heel strike angle improved upon feedback to a degree that, on average, was close to the medication effect on parkinsonian gait. Stimulation at random time points showed intermediate values with non-significant improvements in gait parameters compared to control walks. Closed-loop feedback resulted in a less variable gait pattern compared to control walks. Persons with Parkinson’s disease rated the feedback mode as subjectively useful. Conclusions Our results demonstrate the use of closed-loop feedback can improve parkinsonian gait and suggest such a device could effectively complement the available approaches in the treatment of persons with Parkinson’s disease. Trial Registration: This trial was retrospectively registered in the German Clinical Trials Register (DRKS00038516) on 9 December 2025 .
