Performance-Based Robotic Training in Individuals with Subacute Stroke: Differences between Responders and Non-Responders.

The high variability of upper limb motor recovery with robotic training (RT) in subacute stroke underscores the need to explore differences in responses to RT. We explored differences in baseline characteristics and the RT dose between responders (ΔFugl-Meyer Assessment (FMA) score ≥ 9 points; n = 20) and non-responders (n = 16) in people with subacute stroke (mean [SD] poststroke time at baseline, 54 (26) days, baseline FMA score, 23 (17) points) who underwent 16 RT sessions combined with conventional therapies. Baseline characteristics were compared between groups. During RT sessions, the actual practice time (%), number of movements performed, and total distance covered (cm) in assisted and unassisted modalities were compared between groups. At baseline, participant characteristics and FMA scores did not differ between groups. During the RT, non-responders increased practice time (+15%; p = 0.02), performed more movements (+285; p = 0.004), and covered more distance (+4037 cm; p < 10-3), with no difference between physical modalities. In contrast, responders decreased practice time (-21%; p = 0.01) and performed fewer movements (-338; p = 0.03) in the assisted modality while performing more movements (+328; p < 0.05) and covering a greater distance (+4779 cm; p = 0.01) in unassisted modalities. Despite a large amount of motor practice, motor outcomes did not improve in non-responders compared to responders: the difficulty level in RT may have been too low for them. Future studies should combine robot-based parameters to describe the treatment dose, especially in people with severe-to-moderate arm paresis, to optimize the RT and improve the recovery prognosis.

Robot-assisted therapy for upper limb paresis after stroke: Use of robotic algorithms in advanced practice.

BACKGROUND: Rehabilitation of stroke-related upper limb paresis is a major public health issue. OBJECTIVE: Robotic systems have been developed to facilitate neurorehabilitation by providing key elements required to stimulate brain plasticity and motor recovery, namely repetitive, intensive, adaptative training with feedback. Although the positive effect of robot-assisted therapy on motor impairments has been well demonstrated, the effect on functional capacity is less certain. METHOD: This narrative review outlines the principles of robot-assisted therapy for the rehabilitation of post-stroke upper limb paresis. RESULTS: A paradigm is proposed to promote not only recovery of impairment but also function. CONCLUSION: Further studies that would integrate some principles of the paradigm described in this paper are needed.

Relation between Cortical Activation and Effort during Robot-Mediated Walking in Healthy People: A Functional Near-Infrared Spectroscopy Neuroimaging Study (fNIRS).

Force and effort are important components of a motor task that can impact rehabilitation effectiveness. However, few studies have evaluated the impact of these factors on cortical activation during gait. The purpose of the study was to investigate the relation between cortical activation and effort required during exoskeleton-mediated gait at different levels of physical assistance in healthy individuals. Twenty-four healthy participants walked 10 m with an exoskeleton that provided four levels of assistance: 100%, 50%, 0%, and 25% resistance. Functional near-infrared spectroscopy (fNIRS) was used to measure cerebral flow dynamics with a 20-channel (plus two reference channels) device that covered most cortical motor regions bilaterally. We measured changes in oxyhemoglobin (HbO2) and deoxyhemoglobin (HbR). According to HbO2 levels, cortical activation only differed slightly between the assisted conditions and rest. In contrast, bilateral and widespread cortical activation occurred during the two unassisted conditions (somatosensory, somatosensory association, primary motor, premotor, and supplementary motor cortices). A similar pattern was seen for HbR levels, with a smaller number of significant channels than for HbO2. These results confirmed the hypothesis that there is a relation between cortical activation and level of effort during gait. This finding should help to optimize neurological rehabilitation strategies to drive neuroplasticity.

Using Robot-Based Variables during Upper Limb Robot-Assisted Training in Subacute Stroke Patients to Quantify Treatment Dose.

In post-stroke motor rehabilitation, treatment dose description is estimated approximately. The aim of this retrospective study was to quantify the treatment dose using robot-measured variables during robot-assisted training in patients with subacute stroke. Thirty-six patients performed fifteen 60 min sessions (Session 1−Session 15) of planar, target-directed movements in addition to occupational therapy over 4 (SD 2) weeks. Fugl−Meyer Assessment (FMA) was carried out pre- and post-treatment. The actual time practiced (percentage of a 60 min session), the number of repeated movements, and the total distance traveled were analyzed across sessions for each training modality: assist as needed, unassisted, and against resistance. The FMA score improved post-treatment by 11 (10) points (Session 1 vs. Session 15, p < 0.001). In Session 6, all modalities pooled, the number of repeated movements increased by 129 (252) (vs. Session 1, p = 0.043), the total distance traveled increased by 1743 (3345) cm (vs. Session 1, p = 0.045), and the actual time practiced remained unchanged. In Session 15, the actual time practiced showed changes only in the assist-as-needed modality: −13 (23) % (vs. Session 1, p = 0.013). This description of changes in quantitative-practice-related variables when using different robotic training modalities provides comprehensive information related to the treatment dose in rehabilitation. The treatment dose intensity may be enhanced by increasing both the number of movements and the motor difficulty of performing each movement.

Impact of Dose of Combined Conventional and Robotic Therapy on Upper Limb Motor Impairments and Costs in Subacute Stroke Patients: A Retrospective Study.

INTRODUCTION: Robot-based training integrated into usual care might optimize therapy productivity and increase treatment dose. This retrospective study compared two doses of an upper limb rehabilitation program combining robot-assisted therapy and occupational therapy on motor recovery and costs after stroke. METHODS: Thirty-six subacute stroke patients [Fugl-Meyer Assessment (FMA) score 32 ± 12 points; mean ± SD] underwent a combined program of 29 ± 3 sessions of robot-assisted therapy and occupational therapy. Scheduled session time for the higher dose group (HG) was 90 min (two 45-min sessions; n = 14) and for the lower dose group (LG) was 60 min (two 30-min sessions; n = 22). Pre-/post-treatment change in FMA score (ΔFMA, %), actual active time (min), number of movements and number of movements per minute per robot-assisted therapy session were compared between groups. The costs of the combined programs were also analyzed. RESULTS: ΔFMA did not differ significantly between groups; the HG improved by 16 ± 13 % and the LG by 11 ± 8%. A between-group difference was found for actual active time (p = 1.06E-13) and number of movements (p = 4.42E-2) but not for number of movements per minute during robot-assisted therapy: the HG performed 1,023 ± 344 movements over 36 ± 3 min and the LG performed 796 ± 301 movements over 29 ± 1 min. Both groups performed 28 movements per minute. The combined program cost was €2017 and €1162 for HG and LG, respectively. CONCLUSIONS: Similar motor improvements were observed following two doses of movement-based training. The reduction in scheduled session time did not affect the intensity of the practice and met economic constraints.

Adapted physical activity in community-dwelling adults with neurological disorders: design and outcomes of a fitness-center based program.

PURPOSE: To describe and evaluate an adapted physical activity-based program (APA-program) designed for community-dwelling patients with neurological diseases in local fitness centers. MATERIALS AND METHODS: The APA-program consisted of individual and group activities supervised by an adapted physical activity (APA) instructor twice a week for 6 months. Clinical evaluations before and after the APA-program included strength tests on gym machines, the six-minute walk test (6MWT), the single-leg stance test, and the Short Form-36 (SF-36). RESULTS: Between January 2017 and May 2019, 79 individuals participated in the APA-program (33 women, mean age 59 ± 14 years, 47 stroke, 13 multiple sclerosis, and 19 other neurological diseases). All physical outcomes improved significantly: upper body strength increased by 49%, lower body strength by 37%, 6MWT by 22%, single-leg stance time by 86%, and SF-36 Mental and Physical scores by 23%. Sixty-eight percent of participants completed the 6-month program and 83% of completers then purchased a one-year subscription in the fitness center. CONCLUSIONS: The 6-month APA-program improved participant's physical abilities and quality of life. More than half of participants decided to subscribe personally to the fitness center at the end of the program, supporting the development of community APA-programs for people with chronic neurological diseases, in collaboration with rehabilitation hospitals.Implications for rehabilitationA physical activity-based-program implemented in a fitness center is effective in improving physical fitness and quality of life in people suffering from neurological disorders such as stroke and multiple sclerosis.Close coordination between rehabilitation hospitals and local fitness centers, training by skilled adapted physical activity instructors and group activities are critical determinants of successful participation.

Comparison of active-assisted and active-unassisted robot-mediated upper limb therapy in subacute stroke.

BACKGROUND: Upper-limb robot-mediated therapy is usually carried out in active-assisted mode because it enables performance of many movements. However, assistance may reduce the patient's own efforts which could limit motor recovery. OBJECTIVE: The aim of this study was to compare the effects of active-assisted and active-unassisted robotic interactions on motor recovery in subacute stroke patients with moderate hemiparesis. METHODS: Fourteen patients underwent a 6-week combined upper limb program of usual therapy and robotic therapy using either the active-unassisted (n = 8) or active-assisted (n = 6) modes. In the active-assisted group, assistance was only provided for the first 3 weeks (1st period) and was then switched off for the remaining 3 weeks (2nd period). The Fugl-Meyer Assessment (FMA) was carried out pre- and post-treatment. The mean number of movements performed and the mean working distance during the 1st and 2nd periods were compared between groups. RESULTS: FMA score improved post-treatment in both groups with no between-group differences: active-assisted group: +8±6 pts vs active-unassisted group: +10±6 pts (ns). Between the 1st and 2nd periods, there was a statistical trend towards an improvement in the number of movements performed (p = 0.06) in the active-unassisted group (526±253 to 783±434, p = 0.06) but not in the active-assisted group (882±211 to 880±297, ns). Another trend of improvement was found for the working distance in the active-unassisted group (8.7±4.5 to 9.9±4.7, p = 0.09) but not in the active-assisted group (14.0±0 to 13.5±1.1, ns). CONCLUSIONS: The superiority of the non-assistive over assistive robotic modes has not been demonstrated. However, the non-assistive mode did not appear to reduce motor recovery in this population, despite the performance of fewer movements on shorter working distance compared with the group who had assistance. It seems that the requirement of effort could be a determinant factor for recovery in neurorehabilitation however further well-design studies are needed to fully understand this phenomenon.

Robot-Assisted Therapy in Upper Extremity Hemiparesis: Overview of an Evidence-Based Approach.

Robot-mediated therapy is an innovative form of rehabilitation that enables highly repetitive, intensive, adaptive, and quantifiable physical training. It has been increasingly used to restore loss of motor function, mainly in stroke survivors suffering from an upper limb paresis. Multiple studies collated in a growing number of review articles showed the positive effects on motor impairment, less clearly on functional limitations. After describing the current status of robotic therapy after upper limb paresis due to stroke, this overview addresses basic principles related to robotic therapy applied to upper limb paresis. We demonstrate how this innovation is an evidence-based approach in that it meets both the improved clinical and more fundamental knowledge-base about regaining effective motor function after stroke and the need of more objective, flexible and controlled therapeutic paradigms.

Can robot-based measurements improve prediction of motor performance after robot-assisted upper-limb rehabilitation in patients with moderate-to-severe sub-acute stroke?

PURPOSE: Patients with moderate-to-severe stroke-related upper limb impairment can benefit from repetitive robot-assisted training. However, predicting motor performance in these patients from baseline measurements, including robot-based parameters would help clinicians to provide optimal treatments for each individual. METHODS: Forty-six patients with sub-acute stroke underwent a 16-session upper limb rehabilitation combining usual care and robotic therapy. Motor outcomes (Fugl-Meyer Assessment Upper Extremity (FMA) score) were retrospectively analysed and potential predictors of motor outcome (including baseline FMA scores, kinematics and number of repetitions performed in the first session etc.) were determined. RESULTS: The 16-sessions upper limb combined training program led to significantly improved clinical outcomes (gains of 13.8±11.2 for total FMA score and 7.3±6.7 for FMA Shoulder/Elbow score). For the prediction model, time since stroke poorly explained the FMA total score (R2 < 35%). The model however found that time since stroke and initial value of FMA Shoulder/Elbow score were predictors of the FMA Shoulder/Elbow score: (R2 = 59.6%). CONCLUSION: This study found that clinical prediction of motor outcomes after moderate-to-severe upper-limb paresis is limited. However, initial proximal motor impairment severity predicted proximal motor performance. The value of baselines kinematics and of the number of repeated movements at initiation in the prediction would need further studies.

Anatomical predictors of successful prism adaptation in chronic visual neglect.

Visual neglect is a frequent and disabling consequence of right hemisphere damage. Previous work demonstrated a probable role of posterior callosal dysfunction in the chronic persistence of neglect signs. Prism adaptation is a non-invasive and convenient technique to rehabilitate chronic visual neglect, but it is not effective in all patients. Here we aimed to assess the hypothesis that prism adaptation improves left neglect by facilitating compensation through the contribution of the left, undamaged hemisphere. We assessed the relationship between prism adaptation effects, cortical thickness and white matter integrity in a group of 14 patients with unilateral right-hemisphere strokes and chronic visual neglect. Results showed that patients who benefitted from prism adaptation had thicker cortex in temporo-parietal, prefrontal and cingulate areas of the left, undamaged hemisphere. Additionally, these patients had a higher fractional anisotropy value in the body and genu of the corpus callosum. Results from normal controls show that these callosal regions connect temporo-parietal, sensorimotor and prefrontal areas. Finally, shorter time intervals from the stroke tended to improve patients' response to prism adaptation. We concluded that prism adaptation may improve left visual neglect by promoting the contribution of the left hemisphere to neglect compensation. These results support current hypotheses on the role of the healthy hemisphere in the compensation for stroke-induced, chronic neuropsychological deficits, and suggest that prism adaptation can foster this role by exploiting sensorimotor/prefrontal circuits, especially when applied at early stages post-stroke.

Evolution of upper limb kinematics four years after subacute robot-assisted rehabilitation in stroke patients.

Purpose: To assess functional status and robot-based kinematic measures four years after subacute robot-assisted rehabilitation in hemiparesis. Materials and methods: Twenty-two patients with stroke-induced hemiparesis underwent a ≥3-month upper limb combined program of robot-assisted and occupational therapy from two months post-stroke, and received community-based therapy after discharge. Four years later, 19 (86%) participated in this follow-up study. Assessments 2, 5 and 54 months post-stroke included Fugl-Meyer (FM), Modified Frenchay Scale (MFS, at Month 54) and robot-based kinematic measures of targeting tasks in three directions, north, paretic and non-paretic: distance covered, velocity, accuracy (root mean square (RMS) error from straight line) and smoothness (number of velocity peaks; upward changes in accuracy and smoothness represent worsening). Analysis was stratified by FM score at two months: ≥17 (Group 1) or <17 (Group 2). Correlation between impairment (FM) and function (MFS) was explored at 54 months. Results: FM scores were stable from 5 to 54 months (+1[-2;4], median [1st; 3rd quartiles], ns). Kinematic changes (three directions pooled) were: distance -1[-17;2]% (ns); velocity, -8[-32;28]% (ns); accuracy, +6[-13;98]% (ns); smoothness, +44[-6;126]% (p < 0.05). Group 2 showed decline vs. Group 1 (p < 0.001) in FM (Group 1, +3[1;5], p < 0.01; Group 2, -7[-11;-1], ns) and accuracy (Group 1, -3[-27;38]%, ns; Group 2, +29[17;140]%, p < 0.001). At 54 months, FM and MFS were highly correlated (Pearson's rho = 0.89; p < 0.001). Conclusions: While impairment appeared stable four years after robot-assisted upper limb training during subacute post-stroke phase, movement kinematics deteriorated despite community-based therapy, especially in more severely impaired patients. Trial registration: EudraCT 2016-005121-36. Registration: 2016-12-20. Date of enrolment of the first participant to the trial: 2009-11-24.

Common brain networks for distinct deficits in visual neglect. A combined structural and tractography MRI approach.

Visual neglect is a heterogeneous, multi-component syndrome resulting from right hemisphere damage. Neglect patients do not pay attention to events occurring on their left side, and have a poor functional outcome. The intra-hemispheric location of lesions producing neglect is debated, because studies using different methods reported different locations in the grey matter and in the white matter of the right hemisphere. These reported locations show various patterns of overlapping with the fronto-parietal attention networks demonstrated by functional neuroimaging. We explored the anatomical correlates of neglect patients' performance on distinct tests of neglect. For the first time in neglect anatomy studies, we individually assessed 25 patients with subacute strokes in the right hemisphere, by using a combined structural and diffusion tensor deterministic tractography approach, with separate analyses for each neglect test. The results revealed that lesions in nodes of the ventral attention network (angular and supramarginal gyri) were selectively associated with deficits in performance on all the tests used; damage to other structures correlated with impaired performance on specific tests, such as the bells test (middle and inferior frontal gyri), or the reading test (temporal regions). Importantly, however, white matter damage proved crucial in producing neglect-related deficits. Voxel-based lesion-symptom mapping (VLSM) and tractography consistently revealed that damage to the ventral branch of the superior longitudinal fasciculus (SLF III) and to the inferior fronto-occipital fasciculus (IFOF) predicted pathological scores on line bisection/drawing copy and on the bells test, respectively. Moreover, damage to distinct sectors of SLF III, or combined SLF/IFOF damage, gave rise to different performance profiles. Our results indicate that both grey and white matter lesion analysis must be taken into account to determine the neural correlates of neglect-related deficits. They also suggest that damage to distinct portions of white matter tracts may give rise to distinct clinical signs of neglect, presumably by inducing dysfunction of partly overlapping, but distinct networks.

Component deficits of visual neglect: "Magnetic" attraction of attention vs. impaired spatial working memory.

Visual neglect is a disabling consequence of right hemisphere damage, whereby patients fail to detect left-sided objects. Its precise mechanisms are debated, but there is some consensus that distinct component deficits may variously associate and interact in different patients. Here we used a touch-screen based procedure to study two putative component deficits of neglect, rightward "magnetic" attraction of attention and impaired spatial working memory, in a group of 47 right brain-damaged patients, of whom 33 had signs of left neglect. Patients performed a visual search task on three distinct conditions, whereby touched targets could (1) be tagged, (2) disappear or (3) show no change. Magnetic attraction of attention was defined as more left neglect on the tag condition than on the disappear condition, where right-sided disappeared targets could not capture patients' attention. Impaired spatial working memory should instead produce more neglect on the no change condition, where no external cue indicated that a target had already been explored, than on the tag condition. Using a specifically developed analysis algorithm, we identified significant differences of performance between the critical conditions. Neglect patients as a group performed better on the disappear condition than on the no change condition and also better in the tag condition comparing with the no change condition. No difference was found between the tag condition and the disappear condition. Some of our neglect patients had dissociated patterns of performance, with predominant magnetic attraction or impaired spatial working memory. Anatomical results issued from both grey matter analysis and fiber tracking were consistent with the typical patterns of fronto-parietal and occipito-frontal disconnection in neglect, but did not identify lesional patterns specifically associated with one or another deficit, thus suggesting the possible co-localization of attentional and working memory processes in fronto-parietal networks. These findings give support to the hypothesis of the co-occurrence of distinct cognitive deficits in visual neglect and stress the necessity of multi-component models of visuospatial disorders.

Combining Upper Limb Robotic Rehabilitation with Other Therapeutic Approaches after Stroke: Current Status, Rationale, and Challenges.

A better understanding of the neural substrates that underlie motor recovery after stroke has led to the development of innovative rehabilitation strategies and tools that incorporate key elements of motor skill relearning, that is, intensive motor training involving goal-oriented repeated movements. Robotic devices for the upper limb are increasingly used in rehabilitation. Studies have demonstrated the effectiveness of these devices in reducing motor impairments, but less so for the improvement of upper limb function. Other studies have begun to investigate the benefits of combined approaches that target muscle function (functional electrical stimulation and botulinum toxin injections), modulate neural activity (noninvasive brain stimulation), and enhance motivation (virtual reality) in an attempt to potentialize the benefits of robot-mediated training. The aim of this paper is to overview the current status of such combined treatments and to analyze the rationale behind them.

Pattern of improvement in upper limb pointing task kinematics after a 3-month training program with robotic assistance in stroke.

BACKGROUND: When exploring changes in upper limb kinematics and motor impairment associated with motor recovery in subacute post stroke during intensive therapies involving robot-assisted training, it is not known whether trained joints improve before non-trained joints and whether target reaching capacity improves before movement accuracy. METHODS: Twenty-two subacute stroke patients (mean delay post-stroke at program onset 63 ± 29 days, M2) underwent 50 ± 17 (mean ± SD) 45-min sessions of robot-assisted (InMotion™) shoulder/elbow training over 3 months, in addition to conventional occupational therapy. Monthly evaluations (M2 to M5) included Fugl-Meyer Assessment (FM), with subscores per joint, and four robot-based kinematic measures: mean target distance covered, mean velocity, direction accuracy (inverse of root mean square error from straight line) and movement smoothness (inverse of mean number of zero-crossings in the velocity profile). We assessed delays to reach statistically significant improvement for each outcome measure. RESULTS: At M5, all clinical and kinematic parameters had markedly improved: Fugl-Meyer, +65% (median); distance covered, +87%; mean velocity, +101%; accuracy, +134%; and smoothness, +96%. Delays to reach statistical significance were M3 for the shoulder/elbow Fugl-Meyer subscore (+43%), M4 for the hand (+80%) and M5 for the wrist (+133%) subscores. For kinematic parameters, delays to significant improvements were M3 for distance (+68%), velocity (+65%) and smoothness (+50%), and M5 for accuracy (+134%). CONCLUSIONS: An intensive rehabilitation program combining robot-assisted shoulder/elbow training and conventional occupational therapy was associated with improvement in shoulder and elbow movements first, which suggests focal behavior-related brain plasticity. Findings also suggested that recovery of movement quantity related parameters (range of motion, velocity and smoothness) might precede that of movement quality (accuracy). TRIAL REGISTRATION: EudraCT 2016-005121-36 . Date of Registration: 2016-12-20. Date of enrolment of the first participant to the trial: 2009-11-24 (retrospective data).

Cortico-thalamic disconnection in a patient with supernumerary phantom limb.

Supernumerary phantom limb (SPL) designates the experience of an illusory additional limb occurring after brain damage. Functional neuroimaging during SPL movements documented increased response in the ipsilesional supplementary motor area (SMA), premotor cortex (PMC), thalamus and caudate. This suggested that motor circuits are important for bodily related cognition, but anatomical evidence is sparse. Here, we tested this hypothesis by studying an extremely rare patient with chronic SPL, still present 3 years after a vascular stroke affecting cortical and subcortical right-hemisphere structures. Anatomical analysis included an advanced in vivo reconstruction of white matter tracts using diffusion-based spherical deconvolution. This reconstruction demonstrated a massive and relatively selective disconnection between anatomically preserved SMA/PMC and the thalamus. Our results provide strong anatomical support for the hypothesis that cortico-thalamic loops involving motor-related circuits are crucial to integrate sensorimotor processing with bodily self-awareness.

Upper limb robotics applied to neurorehabilitation: An overview of clinical practice.

BACKGROUND: During the last two decades, extensive interaction between clinicians and engineers has led to the development of systems that stimulate neural plasticity to optimize motor recovery after neurological lesions. This has resulted in the expansion of the field of robotics for rehabilitation. Studies in patients with stroke-related upper-limb paresis have shown that robotic rehabilitation can improve motor capacity. However, few other applications have been evaluated (e.g. tremor, peripheral nerve injuries or other neurological diseases). PURPOSE: This paper presents an overview of the current use of upper limb robotic systems for neurorehabilitation, and highlights the rationale behind their use for the assessment and treatment of common neurological disorders. CONCLUSIONS: Rehabilitation robots are little integrated in clinical practice, except after stroke. Although few studies have been carried out to evaluate their effectiveness, evidence from the neurosciences and indications from pilot studies suggests that upper limb robotic rehabilitation can be applied safely in various other neurological conditions. Rehabilitation robots provide an intensity, quality and dose of treatment that exceeds therapist-mediated rehabilitation. Moreover, the use of force fields, multi-sensory environments, feedback etc. renders such rehabilitation engaging and motivating. Future studies should evaluate the effectiveness of rehabilitation robots in neurological pathologies other than stroke.

Does assist-as-needed upper limb robotic therapy promote participation in repetitive activity-based motor training in sub-acute stroke patients with severe paresis?

BACKGROUND: Repetitive, active movement-based training promotes brain plasticity and motor recovery after stroke. Robotic therapy provides highly repetitive therapy that reduces motor impairment. However, the effect of assist-as-needed algorithms on patient participation and movement quality is not known. OBJECTIVE: To analyze patient participation and motor performance during highly repetitive assist-as-needed upper limb robotic therapy in a retrospective study. METHODS: Sixteen patients with sub-acute stroke carried out a 16-session upper limb robotic training program combined with usual care. The Fugl-Meyer Assessment (FMA) score was evaluated pre and post training. Robotic assistance parameters and Performance measures were compared within and across sessions. RESULTS: Robotic assistance did not change within-session and decreased between sessions during the training program. Motor performance did not decrease within-session and improved between sessions. Velocity-related assistance parameters improved more quickly than accuracy-related parameters. CONCLUSIONS: An assist-as-needed-based upper limb robotic training provided intense and repetitive rehabilitation and promoted patient participation and motor performance, facilitating motor recovery.