The study demonstrated clinical and neurophysiological changes in response to the robotic-exoskeleton (19) training compared to the conventional-rehabilitation. Clinical-scales showed improvement in both RG and CG, however, increased cortical-excitability in the ipsilesional-hemisphere was shown only in RG with the appearance of MEPs in the ipsilesional-hemisphere post-therapy in patients. The improvement in RMT in the ipsilesional-hemisphere showed a trend of normalization over the intervention and was also correlated with sensorimotor function improvement.
4.1 Comparison of Clinical-scales of Robotic-therapy group with control-group
The robotic-therapy was effective in releasing spasticity at the wrist joint with ~26% (p=0.03) improvement over only ~14% in CG. The regain in normal muscle tone is considered as a predictor of recovery or the first step in recovery (29) followed by an increase in muscle strength and improvement in functional movements. Both groups showed significant improvement of AROM, RG showed significantly (p=0.02) higher improvement of 130% over 47% in CG (Table-2).
FMU/L, stroke-specific scale, is the most reliable measure of sensorimotor functionalities of the whole arm (30). RG established significantly higher improvements in FMU/L of ~40% over ~21% in CG (p=0.039). For FMW/H distal-component, both groups showed significant improvement, where RG showed significantly higher improvement of ~72% compared to only 32% in CG (p=0.012) possibly because of intensive and repetitive training of wrist and MCP. However, RG did not show significantly higher improvement in FMS/E, as expected as the intervention was not focused on the proximal component. With contemporary studies showing improvements even in the proximal component in response to distal training, our study too reflected change in proximal joint (FMS/E) post-therapy, probably because of the compensatory muscle activities from proximal-joints (13)(31).
Thus, RG shows an overall increase in sensorimotor ability and functionality as evidenced by the increase in FM, AROM and decrease in spasticity (MAS); thereby, attributing to increased mobility and stability of wrist extension and hand activities like grasping, which in turn, reflected improvements in FMW/H scores. As shown in the studies by Gladstone et al., and Shin et al., (32)(33), a value of 6.6 (FMU/L) reflects the potential Minimally Clinically Important Difference (MCID). In our study, FMU/L (14 on a scale of 66) was found to be higher than the MCID values for all 12 patients in RG and 7/11 patients in CG. The Hand Mentor Pro, which rotates wrist with MCP placed at a constant angle with respect to the wrist, lacks flexion (grasp) and extension (release) of MCP, patient-centric ROM and speed, reported improvement in 99 patients with stroke with FMW/H being 5.6 (FMU/L 10.33 in combination with a home exercise program (which alone reported FMW/H 4.9 and FMU/L 9.3). The HWARD (31) also showed an improvement in FMW/H (~4) with sensorimotor cortex laterality index representing a shift in interhemispheric balance over time from the contralateral to ipsilateral side and also suggested the use of synchronizing both wrist and MCP-joint movement in grasp and release.
With Constraint-Induced Movement therapy (CIMT), the reported gain was FMU/L ~13 and BI ~13.5, post 3 weeks therapy (34). Moreover, a systematic review and meta-analysis has shown improvement in FMU/L scores Action Research Arm Test (ARAT) scores with an improved control of hand and arm placement as well as improved strength compared to standard therapy post-CIMT in the subacute and chronic stroke populations (35). Few studies have also shown significant increased motor map area via TMS post-CIMT (3)(36). Sawaki et al., showed increased in the TMS motor map area (of EDC muscle in ipsilesion hemisphere in few patients) and clinically relevant improvements in arm motor function that persisted for at least 4 months, however, other TMS parameters like Resting Motor Threshold, Active Motor Threshold, Center of Gravity, silent period did not change over time (37). Use of biofeedback has been another widely explored area, where Doan-Aslan et al. and Zheng et al. has demonstrated an increase in AROM, BI and FMU/L in patients with stroke while using EMG Biofeedback compared to the conventional therapy (38)(39). In our study, an improvement in FMW/H by 3 was observed with intervention therapy compared to the CG, which is consistent with the literarure. Krishnan et al. and Calabrò et al. have attempted to evaluate the effect of active robotic-training on changes in cortical-excitability, using commercially available devices, such as Lokomat robot (lower-limb) (40) and ARMEO (upper-limb) (41). With very sparse literature exploring cortical excitability changes in lower-limb (42) and upper-limb (43), virtual mirror task with feedback demonstrated increased MEP by up to 46.3% (95% CI: 30.4 ~ 80.0) compared with the real mirror task (43).
For the Barthel Index, both groups showed similar (~20%) improvement (p=0.82). Both groups showed significant improvement for BS as well, however, RG showed ~32% improvements compared to only ~20% in CG (Table-2). In the case of the Barthel Index and Brunnstrom Stage, both RG and CG showed a similar improvement as the rehabilitation regimen in the CG group incorporated clinical rehabilitation with a primary focus on the upper extremity deficits with the therapist focusing on the distal limb and overall recovery along with customizing the patient’s goals directly and training compensatory and functional movement strategies that consequently resulted in equal gains in independence and patients’ goals as in RG. Also, in the future, substantial consideration can be given to Barthel Index scores by introducing kinematic analysis of speed, accuracy, and precision of movement and BI-based patient perception scales like self-perceived difficulty scale and ability scale for better quantitative measurement.
4.2 Comparison of Cortical-excitability of Robotic-therapy with the control-group
Cortical-excitability in pre-therapy measurements was found to be lower in patients with stroke as observed by higher RMT and lower MEP, same as reported in (22–26). In some patients due to low cortical-excitability, MEP is not recordable even after delivering TMS stimuli at the highest possible intensity at 100% (22–26). In those patients with no MEP recorded, RMT is taken as a value of 100, as suggested in the literature (27,28). Though a subset of patients with stroke with affected corticospinal tract integrity that does not demonstrate MEP with the highest stimulation intensity, taking RMT as 100% could affect the decrease in RMT post-therapy in the RG group. However, critical studies like Hendrics et al., and Jong et al., have established MEP as a sensitive and valid prognostic marker of motor recovery after stroke (44–46).
4.2.1 Ipsilesional and Contralesional-hemisphere changes
With the decrease in RMT, RG showed ~16% improvement as compared to ~4% improvement in CG (p=0.037). Interestingly, RG showed a significantly (p=0.048) higher increase in MEP-amplitude post-therapy with an increase of ~140% (mean=54.9 µv), whereas CG showed no such improvement. Cortical-excitability measures are used as an objective investigative tool to measure the treatment responsiveness and prognostication as it provides insights into membrane-excitability of neurons, conduction, and functional integrity of the corticospinal tract and neuromuscular-junctions (47). A decrease in RMT and increase in MEP amplitude in the ipsilesional- hemisphere, demonstrated in the RG and not in CG, might be related to the increase in cortical-excitability (48). It might be interpreted that recovery of motor function could most likely be a consequence of plastic reorganization and use-dependent plasticity (48). Cortical-excitability and corticospinal tract integrity have also been shown to be correlated with functional recovery potential in patients with chronic stroke (23) and exoskeleton training appears to be beneficial in activating the ipsilesional-hemisphere for our patient cohort (chronicity 13.8±9.1 months). Activation of ipsilesional-hemisphere could indicate either vicariation of the loss of neural circuits or unmasking of preexisting synapses or recruitment of perilesional areas in ipsilesional-hemisphere or exploitation of the preserved functional recovery reservoir in ipsilesional-hemisphere (27,49–51).
In the contralesional-hemisphere, MEP-amplitude showed a considerable decrease in both groups, though not significant, RG evidencing a decrease of ~30% (mean=151.03 µv) and CG a decrease of only ~7% (mean=14.8 µv) with no significant differences (p=0.51) (Table-2). A ~30% decrease in MEP-amplitude in contralesional-hemisphere over the duration of intervention might represent a decrease in cortical-excitability (49)(50), however, is difficult to comment on it at this stage due to the small sample size and needs to be further evaluated in a larger cohort.
The potential clinical effectiveness harnessed by the neuro-rehabilitation robots has also been shown by few studies in terms of subjective clinical scales or questionnaires or EMG parameters which might not be sufficient to assess cortical reorganization (9)(11)(7,13–15)(52–56). However, the mechanism of entrainment of neuroplasticity followed by a stroke that favors motor learning and functional recovery is still unclear (57). Despite recognizing that the corticospinal tract plays a critical role in recovery potential, cortical reorganization, functional improvement in stroke, and as well as better track clinical progression, the changes in these measures evaluating effects due to intervention are usually limited to the studies involving brain stimulation protocols. Examples are repetitive TMS, Transcranial Direct Current Stimulation (tDCS) (58),(59), etc. or in a combination of brain-stimulation with other neuro-rehabilitation strategies like CIMT (60) or mirror-therapy (61) or training (62),(63). Hence, only limited studies are available assessing for these measures unveiling objective changes using robotic-therapy as a rehabilitation intervention (58–66).
Though the study using the device HWARD provided seminal evidence of reorganization of brain (via fMRI), as well as motor function in response to the robotic-therapy, no direct comparison can be made with our study as different modalities - TMS and fMRI was used to measure different neurophysiological aspects (31). Juan et al., correlated results by these modalities and presented that larger fMRI activation likelihood and motor cortical excitability in the ipsilesional primary motor area were related to improved motor performance (67).
4.2.2 Specific five-patients in RG
A very critical outcome of the therapy was that in RG, MEP was evoked in ipsilesional-hemisphere only for 4/12 patients at the pre-therapy measurements; whereas, MEP was later evoked for 9/12 patients after robotic-therapy. However, in CG, MEP was evoked only for 5/11 patients and was later evoked for 6/11 patients at post-therapy. Considering these five specific patients in RG who did not evoke MEP at pre-therapy and later evoked MEP (mean=136.6±38.4 µv), with a decrease of stimulation intensity in ipsilesional-hemisphere by almost 27% and substantial improvement in the value of clinical-scales (FMW/H by 7.8±2.38, BI by 22±11.72, AROM by 220±2.730). These changes were relatively much higher than the changes in patients who had MEP evoked at pre-therapy measures. The appearance of MEP in five patients after 4 weeks of robotic intervention is a crucial outcome and represents that the robotic-therapy might have the potential of facilitating clinically relevant reorganization of the brain-based on use-dependent plasticity. The observed increase in cortical-excitability and normalization of TMS neurophysiological makers on the ipsilesional-side was also observed to be accompanied by recovery of hand-function, as observed by sensorimotor and functional recovery (by clinical-scales FMW/H, BI & AROM).
4.2.3 Inter-hemispheric differences and asymmetries
The diaschisis between ipsilesional-areas and intact neuronal-networks of contralesional-areas may disturb the cortical-excitability and connectivity patterns of connected, remote, or primary-motor areas of contralesional-hemisphere (via transcallosal-fibers). The effect of robotic-exoskeleton training on cortical-excitability of both hemispheres might be attributed to remodeling of the bilateral primary-motor areas in RG (time*sides p=0.049, F =4.08) which is not shown in CG (time*sides p=0.06, F=3.68).
For cortical excitability to be increased in ipsilesional-hemisphere for patients with stroke, the ipsilesional-RMT should be decreased from pre-to-post-therapy and hence, RMTasymm (RMT Ipsilesional/RMT contralesional) should decrease to approach normalization (27). Significant differences were observed between the groups when TMS-neurophysiological changes over the intervention were expressed in terms of the interhemispheric-asymmetry ratio RMTasymm might be a representative of a trend (p=0.028) towards the normalization of asymmetry of TMS-measures in RG in response to exoskeleton-training than CG. Normalization might indicate the recruitment of peri-lesional areas in the ipsilesional-hemisphere or exploitation of the preserved functional-recovery reservoir in the ipsilesional-hemisphere (27,49–51).
4.3 TMS neurophysiological improvement correlating the motor-outcome of both groups
The amount of change in TMS neurophysiological measures of corticomotor-pathways (∆RMTipsi and ∆RMTasymm-ratio) were found to be associated with the amount of improvement in functional motor-outcome during the rehabilitation of the distal-part of the upper-limb (∆FMW/H) (Figure-3). These parameters were significantly different for RG and CG (∆RMTipsi p=0.0235, ∆RMTasymm-ratio p=0.028 and ∆FMW/H p=0.012). An improvement (decrease) in motor-threshold tend to show greater increases with clinical-outcome and was found to have positive correlation with ∆FMW/H in RG (∆RMTipsi r=0.64, p=0.022) and not in (CG r=0.47, p=0.13) and ∆RMTasymm-ratio (r=0.6, p=0.03) and not in CG (r=0.29, p=0.38) (Figure-3). The improvement (decrease) in RMT, could be associated with recovery of motor function as suggested by (23). This might be most likely due to increased cortical-excitability of preserved motor-pathways as shown in earlier studies in sub-acute and chronic stroke, demonstrating the correlation of improvement in TMS neurophysiological measures with functional improvement (27), (68), (69). These neurophysiological-measures were obtained specifically from the cortical representation of EDC muscle, a clinically affected muscle, with a specific function which was involved in training with a robotic-exoskeleton, whereas, most clinical measures do not necessarily require a particular muscle group and measures motor-function in a broader sense.
Also, these neurophysiological-parameters individually establishes as a good predictor of functional rehabilitation-outcome of hand (∆FMW/H) in RG, indicating that changes in cortical-excitability of ipsilesional-hemisphere might be used to predict the clinical-outcome, hence, emerging as critical recovery parameters to be considered and evaluated in future with a larger data-samples. These might be the plasticity markers predicting the responsiveness of chronic post-stroke patients (41).
4.4 Changes due to the device
The exoskeleton training in RG induced an evident modulation in ipsilesional and contralesional-hemispheres. However, changes in CG were found to be limited only to the clinical-scales, and in addition, the changes in neurophysiological parameters were specifically found in the RG. The decrease of RMT and change in RMT asymmetry from distal-muscle was also accompanied by functional markers-FMW/H evidencing sensorimotor-plasticity, functional recovery with task-dependent rehabilitation. Multiple strategies used during intervention to encourage clinically relevant neuroplasticity were to use movement goals that are specific, measurable, achievable, repetitive, and timed (70). It was also supported with maximizing voluntary residual muscle activity (using EMG thresholds) with real-time extrinsic visual performance biofeedback and intrinsic proprioceptive feedback for sensorimotor integration in every cycle of movement as synergy based training approach for maximizing brain reorganization (71)(72).
Since RG and CG had very similar lesion locations and size with all patients having their motor paths affected, indicating increase in cortical excitability might be attributed to the different interventions in the groups. There was a limited number of cases in individual subgroups, i.e. only 2/12 patients from the RG and 2/11 patients from the CG belonged to the subacute stage (3 months–6 months), the majority of patients are chronic. The CG included 5 subcortical and 6 cortical stroke. The RG included 6 subcortical and 6 cortical stroke. Considering threshold for recovery as MCID for FMU/L 6.6 (32)(33), out of 11 patients in CG, total 7 patients (5 subcortical and 2 cortical) exceeded the threshold and all twelve patients in RG exceeded the threshold. Any conclusion on the trend for sub-acute or chronic stroke and the responders or the non-responders to the intervention would be highly presumptive and misleading because of the small sample size. The outcomes provided critical data to plan a multicentric trial with large sample size in the future to systematically investigate the potential of the exoskeleton.
4.5 Limitations and Future Work
Even though the data are promising, the study had few limitations such as small sample size and lack of an activity level measurement like Wolf Motor Function test and Action Research Arm Test, goal-directed or translation-to-home-use measurements, no midterm clinical assessment, and long term follow-up of patients. As most of the patients at our quaternary hospital came from far places across India and it was not possible to follow-up with them once they have left the city. Another limitation was therapist performing both sets of interventions could not be blinded to the group allocation. There are several ways the study can be improved. The sample size can be increased and patient groups can be further subdivided into sub-acute and chronic stages to evaluate any difference in rehabilitation outcomes, with mid-term clinical assessment and long-term follow-up with the double-blinded protocol. Different distal goal-directed and translation to home measures could be included like WMFT or ARAT, Functional Independence Measure, Canadian Occupational Performance Measure or Motor Activity Log, nine-hole pegboard, stroke Impact scale, Interhemispheric Inhibition measures using TMS, etc. The device currently is in the prototype stage with clinical validation, thus the BIOPAC EMG system was used in data acquisition for research and validation. In the future, this will be easily replaced by an MYOWARE or an in-house build EMG amplifier. The device will have an LCD touch screen for settings and feedback. These features will make the system more aesthetic, compact, and accessible. Once the device is optimized in terms of weight, aesthetics and compactness, it can be deployed for home-based rehabilitation in the future. Also, with a minor modification, the device can synchronize wrist extension with finger extension which can be further explored for outcome in patients with stroke.