An open-loop FES system comprises of a manually controlled device, which is operated by a therapist. During the therapy session, the therapist manually administers electrical stimulation to the specific muscles of the patient, using patient-specific predetermined stimulation parameters such as stimulation intensity, time duration, and ON/OFF cycle.

Numerous studies have been conducted utilizing open-loop FES systems for stroke rehabilitation to regain upper limb motor functions. Experimenting with the effectiveness of FES, Nakipoglu Yuzer et al. (17) applied FES (two channels and four surface electrodes) to the spastic muscles of 30 patients (an RCT), and the improvement of clinical scores indicates that FES effectively reduces wrist flexor spasticity. In (18), Makowski et al. showed that FES produces functional hand opening when the patient is relaxed, but it is overpowered by finger flexor coactivation when the patient voluntarily exerts effort to reach/open the hand. For that, their study proves that the amount of hand opening grows significantly (3.2–8.8 cm) when including FES for both reaching and hand opening muscles even in the presence of submaximal or zero effort. Moreover, Meadmore et al. (19) investigated FES of shoulder, elbow, and wrist muscles: five patients underwent 18 sessions and completed FMA (Fugl-Meyer Assessment) and ARAT (Action Research Arm Test) assessments. The study showed an improvement of 4.4, providing evidence that the integration of low-cost hardware with advanced FES controllers can reduce upper limb impairment. Sun et al. (20) reported the FES for upper limb functional activity practice, used by 9 therapists to set 8 sessions activities with 22 stroke patients. Among them, 17 patients showed a session completion rate >90%, demonstrating its capability of delivering high-intensity therapy compared to traditional face-to-face therapy. Also, Niu et al. (21) illustrated a technique for creating FES patterns based on muscle synergies of a normal subject (three patients–adjusted for each participant and task) using a programmable FES device. Followed by 5-day sessions of intervention using synergy-based FES delivery to another three patients. The outcome of the new technology was measured by improvement in FMA scores (28.6% ± 13.7%). In Chou et al. (22), made use of the latter in the design and test of an automated synergy-based FES system to match electrically induced movements to assist residual movements of patients. Results based on changes in FMA scores indicate that the synchronization produced more consistent compound movements with reduced RMS (root mean square) errors under different triggering conditions. Martín-Odriozola et al. (23) developed the Fesia Grasp device used for hand dexterity rehabilitation of a 69-yearold post-ischemic stroke woman. Following their first study (21), Niu et al. (24) conducted a TOT (Task-oriented training) protocol with repeated forward and lateral reaching movements assisted by synergy-based FES on 16 patients, divided into FES (EG) and Sham (CG) groups over 5-days. Findings of higher FMA than Sham indicate efficacy of open-loop FES system in post-stroke rehabilitation. A detailed overview of research studies regarding open-loop FES rehabilitation is provided in Table 1.

According to Hebb's principle “cells that fire together wire together” (72, 73), suggesting that the coordination of cortical and physical activities during the rehabilitation therapies could lead to an effective improvement of the impaired motor function (74–77). Therefore, a more effective approach may be to interface the FES rehabilitation device with an external system that could enhance the simultaneous activation of the motor cortex during the rehabilitation sessions. In this regard, MI-based BCI systems are an optimum choice, which allows the rehabilitation system to perform the required task based on the patient's imagination of intended motion, allowing more active participation of brain throughout the stroke therapy (78).

A BCI-FES rehabilitation system mainly comprises of a BCI unit (containing EEG element), BCI-FES interface component, and FES module. Some BCI-FES systems also incorporate a virtual reality (VR) paradigm as a part of their setup (Figure 3). In such systems, firstly, the patient is provided the VR environment (on screen or the headset) that contains the pre-programmed therapy session of targeted motion e. g., hands extension/flexion (Figure 3A). The patient will be asked to imagine the task execution displayed in the virtual environment. Each task imagination generates a specific EEG signal, which will be acquired and processed by the BCI unit (Figure 3B). Depending on the imagination, the BCI-FES interface generates a trigger command to control the ON/OFF state of FES stimulation and will also control the stimulation parameters (Figure 3C). Lastly, the FES device provides electrical stimulation to the impaired muscles and hence, facilitate performing the required movements (Figure 3D). To make the strategy effective, the user typically undergoes training to establish a connection between their brain signals and specific motor tasks. This training involves practicing mental tasks or visualizing movements to generate distinct brain patterns that the BCI can recognize and translate into commands for the FES device.

BCI-FES systems are widely used for stroke rehabilitation and several randomized controlled trial (RCT) studies have been performed to investigate the efficiency of BCI-FES systems (32–36). In (32), Cincotti et al. performed an (RCT) to restore hand grasping movements. To assess post-stroke motor recovery, the FMA, MRC (Medical Research Council), and ESS (European Stroke Scale) scores were used. The results showed that the group with BCI-FES therapy achieved better motor recovery than the conventional FES group. Likewise, Li et al. (33) targeted stroke survivors with severe upper extremity paralysis. The study compared the efficiency of the BCI-FES system in comparison to the conventional FES system. The result showed a motor imagery task classification accuracy of 77%, along with a substantial improvement in the rehabilitation outcome scores within the BCI-FES group. In (34), Kim et al. accomplished an RCT to investigate the positive influence of the BCI-FES system on the motor recovery of upper extremities in stroke survivors. The measured outcomes validated the enhanced recovery via a BCI-based system compared to physical training. Additionally, in Miao et al. and Chen et al. (35, 36), the clinical application of the BCI-FES stroke rehabilitation system has been proposed to promote and improve upper extremity movements, along with motor activity restoration.

In addition to RCT, different feasibility studies for exploring the applicability of BCI-FES systems have also been carried out (37–45). In Daly et al. (37), Daly et al. performed a pilot study in which they tested a customized BCI-FES system on a stroke survivor having a joint extension problem in the index finger. During the first rehabilitation session, results showed a higher classification accuracy of 97% and 83% for “attempted movement” and “imagined movements” respectively. With every session, the muscle movement was gradually improving and by the end of nine sessions, the finger extension motion was completely recovered. Additionally, Mukaino et al. (38) developed a BCI-controlled neuromuscular electrical stimulator and conducted a case study on a stroke survivor (finger movement) to examine the effectiveness of BCI in stroke therapy. The results indicated that rehabilitation training with a BCI-controlled FES induces cortical plasticity and promotes functional recovery. Apart from customized BCI-FES stroke rehabilitation systems, “RecoveriX from g.tec” is commercially available stroke rehab systems (39). The RecoveriX system classifies the right and left wrist motion intention and is only meant for the wrist dorsiflexion rehabilitation paradigm. Hence, to validate the efficacy of RecoveriX system, Sabathiel et al. (40), Irimia et al. (41), Cho et al. (42), Qiu et al. (43), and Sebastián-Romagosa et al. (44) have conducted experiments on a set of stroke survivors for arm function restoration. Their results showed that the system depicts a classification accuracy of up to 95%. Furthermore, significant improvements of upper limb motor function scores suggest the post-stroke motor recovery. A detailed overview of research studies regarding BCI-FES rehabilitation is provided in Table 2.

EMG provides information on the neural activity of muscles and can detect physical movement intentions. A method has been previously studied to engage a user during FES therapy by triggering the stimulation when a specific level of muscle activity is detected (79–83). In the “EMG triggered FES system”, the EMG signal acts as a switch to trigger the delivery of FES stimulation at a predetermined level when the EMG magnitude reaches a certain threshold. However, this approach only uses the user's muscle activity to trigger FES and has not been conclusively proven advantageous over the open-loop FES method (79–83). Thus, another system, an “EMG controlled FES system” has been adopted that among with an FES trigger, also modulates the FES intensity in proportion to the real-time EMG signal (84).

EMG-controlled FES system mainly comprises of an EMG sensing unit, EMG-FES interface component, and FES module (Figure 4). Certain EMG-FES systems also integrate a virtual reality (VR) component into their configuration. In these systems, the initial phase entails immersing the patient in a VR environment, which can be presented on a screen or through a headset (Figure 4A). This VR environment includes a pre-programmed therapy session focused on specific movements, such as hand extension or flexion. Before the start of a therapy session, the system is calibrated for setting the EMG threshold level and required maximum FES stimulation (varies across subjects). The subject tries to perform the required task (for instance, wrist extension) and the intended motion is physically detected by the EMG sensing unit via analyzing the muscle activity (Figure 4B). The acquired EMG signal is processed by the EMG-FES interface and once the myoelectric activity reaches the pre-defined threshold level, the interface unit sends the trigger command to start the FES (Figure 4C). The applied stimulation activates the targeted muscle (or group of muscles) and helps the subject to achieve the desired motion (Figure 4D). In EMG-FES controlled system, the amount of stimulation does not stay constant and automatically adjusts throughout the therapy sessions proportional to real-time muscle activity.

Shindo et al. (51) performed an RCT to test the efficacy of the myoelectrical controlled electrical stimulator developed by Muraoka (52). The therapy sessions were performed for finger extension rehabilitation (i.e., a functional opening of the hand) which lasted for 3 weeks (5 days/week). The EMG electrodes were placed on the paretic extensor digitorum communis muscles and based on the muscle activities the amount of applied stimulation was controlled. After completion of rehabilitation sessions, pre, and post-performance was evaluated via different clinical score metrics (FMA and ARAT). They found that the EMG-controlled FES was able to induce a greater level of improvement as compared to the control group. In (53, 54), an EMG-controlled FES system, “MeCFES” has been developed by Thorsen's group for upper limb stroke rehabilitation, which was tested via RCT on 11 stroke survivors (55). In the experimental group, the EMG electrodes were placed on wrist and finger extensors and their recorded muscle activity was used to control the applied electrical stimulation for wrist and finger extension. The clinical evaluation was performed through the ARAT, and results showed that the participants treated with MeCFES had a significant improvement in upper limb motor function. In (56), Thorsen's group conducted another RCT in which they tested the MeCFES for task-oriented therapy (TOT). This was the first large RCT (68 stroke survivors) in which multiple rehabilitation centers validated the performance of MeCFES-assisted TOT against standard TOT. In the end, promising results were obtained in terms of MeCFES functioning, and no adverse events were reported in any of the centers. They concluded that MeCFES is a safe and efficient myo-controlled FES system for the motor recovery of upper extremities among stroke survivors. Recently (57), they developed an updated version of MecFES, named “FITFES”, which is wearable and portable in an ambulatory setting and best suitable for TOT applications. Thus far, only a working prototype has been tested on a single subject and no clinical evaluation has been performed. Moreover, Hara et al. (58) investigated the relationship between brain cortical perfusion (BCP) changes in the sensory-motor cortex (SMC) area and arm function improvement. A near-infrared spectroscopy (NIRS) approach was adopted to analyze BCP changes. It was found that EMG-FES rehabilitation improved FMA and GS (grip strength) scores. Also, NIRS showed increased SMC activation during therapy, confirming the functional improvement due to the EMG-FES system. A detailed overview of research studies using EMG-FES rehabilitation is provided in Table 3.

Among open-loop FES systems, the pooled analysis of 3 studies (19, 21, 24) including 17 stroke patients showed a significant increase in FMA score [MD = 5.6, 95% CI (3.77, 7.5), P < 0.001], and the data were found to be homogenous (I² = 0, P = 0.657) (Supplementary Figure S3).

For BCI-controlled FES, the meta-analysis of 6 studies (33–36, 38, 44) with a total of 99 patients exhibited an improvement in FMA score [MD = 5.37, 95% CI (4.2, 6.6), P < 0.001], along with the homogeneity in data (I² = 0, P = 0.198) (Supplementary Figure S3).

Finally, after analyzing the data from 3 EMG-controlled FES studies (51, 56, 58) involving 60 patients, it was found that EMG-FES rehabilitation led to a significant increase in FMA score [MD = 14.14, 95% CI (11.72, 16.6), P < 0.001], and the data were homogenous (I² = 0, P = 0.006) (Supplementary Figure S3).

The meta-analysis of 3 EMG-based FES studies (51, 55, 56) including 49 patients indicated a statistically significant increase in the ARAT score [MD = 11.9, 95% CI (8.8, 14.9), P < 0.001], and the data demonstrated homogeneity (I² = 0, P = 0.534) (Supplementary Figure S4).

Numerous studies did not conduct randomized controlled trials; instead, they just conducted feasibility studies within the stroke population (18–24, 37–45, 58). Such studies included no control group and only performed the rehabilitation protocols on the experimental group. A control group provides a baseline against which the treatment group can be compared. Without a control group, assessing whether any observed changes are greater or different from what would naturally occur without the intervention is challenging. Also, it may be challenging to generalize the study's findings to a broader population or to other settings because there is no comparison to determine whether the effects are consistent across different contexts.

Reported RCTs (17, 32–36, 51, 55, 56) and feasibility studies (19, 21, 23, 24, 38, 42, 44, 58) claimed statistical significance results; however, their sample size is not large enough (lies between 1 and 51 stroke patients in one group). According to Kaptein (94), a conventional RCT requires a group size of at least 64 individuals in each group to obtain statistically significant results. Hence, a small sample size questions the credibility and reliability of studies. It indicates that further investigation or a larger sample size may be needed to establish a more definitive relationship.

Also, there was no mention of the follow-up data to determine whether the improvement was retained or not (17–24, 32–45, 58). The absence of follow-up data in rehabilitation can impede the assessment of long-term outcomes, the identification of relapses, and the ability to make informed decisions about treatment effectiveness and planning. It is crucial for both clinical practice and research to include follow-up assessments to ensure that the benefits of rehabilitation are sustained and optimized over time.

When an individual experiences functional improvement, such as regaining motor skills after a stroke rehabilitation, the brain can reorganize its neural circuits and establish new connections or strengthen existing ones to support improved function (95, 96). These neuronal changes can be determined by different techniques, which mainly include electroencephalography (EEG)/evoked potentials (ERPs), structural and functional magnetic resonance imaging (MRI), and transcranial magnetic stimulation (TMS) (97). Studies (17, 19, 21, 23, 24, 32–36, 38, 42, 44, 51, 55, 56) have shown that the different FES-based rehabilitation causes functional improvement among stroke patients, but none of them has validated their findings by presenting the neuroplasticity outcomes. Thus, it remains uncertain to what extent neuroplasticity has occurred due to open/closed-loop FES rehabilitation.

Hence, it is hard to conclude which specific FES system is best. However, many research studies showed that closed-loop FES is more effective than open-loop FES for motor recovery (32–35, 38, 55). Among closed-loop FES, which system is more efficient (either BCI-FES or EMG-FES) remains unknown, as currently, no RCT has been conducted to directly compare their efficacy in neurorehabilitation. Furthermore, from the clinical implementation point of view, an open-loop FES has been widely used clinically for many years (for stroke rehabilitation), whereas closed-loop FES is mainly applied in the laboratory as a research protocol (especially BCI-FES). As per our knowledge, “RecoveriX from g.tec” (39) is the only commercially available BCI-FES system for stroke rehabilitation, which is also in its initial phases to be adopted by clinicians/therapists.

In (98), the integration of electrical stimulation with robotic arm training resulted in significant improvements in the range of motion for shoulder and elbow movements in subacute stroke survivors, compared to conventional robotic training. Meadmore et al. (99) developed a new rehabilitative system, featuring FES, robotic support, and voluntary effort. The results demonstrated improvements in arm impairment among five stroke survivors. Another study (100) tested an EMG-driven FES-robotic system on 11 chronic stroke survivors to rehabilitate finger, wrist, and elbow movement. Significant improvement in physical functions and arm impairment has been obtained. Qian et al. (101) used the same FES-robotic system on 24 sub-acute stroke survivors, which showed higher motor outcomes at the distal joints than the control group (conventional therapy). Although there are potential advantages in using hybrid FES robotic systems for upper limb rehabilitation, a review study has revealed that only a limited number of hybrid systems have undergone testing with stroke survivors (102). This could be due to challenges associated with integrating both rehabilitation technologies or the absence of integrated platforms that could be user-friendly and easy to set up.

Ambrosini et al. (103) developed a novel hybrid neurorehabilitation system that integrated a passive exoskeleton (named RETRAINER) with an EMG-triggered FES unit. In (103), they tested the feasibility and functionality of the hybrid system in a clinical environment. Later, they performed a pilot study (104) and RCT (105) to test the performance of the developed system for upper limb recovery. The pilot study was implemented on seven post-acute stroke survivors. Preliminary results confirmed that the hybrid FES exoskeleton system can be used for stroke rehabilitation, positively impacting arm functional recovery (104). In (105), an RCT involving 72 stroke survivors validated the performance of a hybrid system compared to advanced conventional therapy (ACT) for task-oriented arm training. The findings showed that the hybrid FES exoskeleton system achieved a significantly better improvement in upper limb functionality.

During the FES-based rehabilitation therapy, the participants started losing interest, and it became difficult for them to maintain the training motivation. This decline in the level of engagement could be attributed to the extended duration of the sessions, the repetitive exercises involved, and the clinical environment in which the rehabilitation took place (106). Therefore, physical therapists increasingly turn to virtual reality (VR) paradigms and incorporate VR into their neurorehabilitation protocols (107). By providing a virtual environment with thrilling, stimulating, and entertaining tasks, VR can keep participants more focused and motivated during rehab exercises, potentially engaging additional neural circuits to restore motor functions more effectively. Hence, the RecoveriX system combines VR with FES and commercially introduced hybrid VR-based BCI-FES stroke rehabilitation systems (39). Different studies (40–44) suggested that the RecoveriX system caused the improvement in upper extremity movements via stroke rehabilitation.

However, as VR is a newly adopted method in neurorehabilitation, initial testing has mostly been performed on small populations. Furthermore, low-quality VR may cause simulator sickness in stroke survivors, thus necessitating high-quality VR that replicates actual environments as realistically as possible. Thus, more research is needed to investigate the practical implementation and feasibility of hybrid VR-based FES systems for neurorehabilitation.