Application of Home-Based Wearable Technologies in Physical Rehabilitation for Stroke: A Scoping Review

Using wearable technologies in the home setting is an emerging option for self-directed rehabilitation. A comprehensive review of its application as a treatment in home-based stroke rehabilitation is lacking. This review aimed to 1) map the interventions that have used wearable technologies in home-based physical rehabilitation for stroke, and 2) provide a synthesis of the effectiveness of wearable technologies as a treatment choice. Electronic databases of the Cochrane Library, MEDLINE, CINAHL, and Web of Science were systematically searched for work published from their inception to February 2022. This scoping review adopted Arksey and O’Malley’s framework in the study procedure. Two independent reviewers screened and selected the studies. Twenty-seven were selected in this review. These studies were summarized descriptively, and the level of evidence was assessed. This review identified that most research focused on improving the hemiparetic upper limb (UL) function and a lack of studies applying wearable technologies in home-based lower limb (LL) rehabilitation. Virtual reality (VR), stimulation-based training, robotic therapy, and activity trackers are the interventions identified that apply wearable technologies. Among the UL interventions, “strong” evidence was found to support stimulation-based training, “moderate” evidence for activity trackers, “limited” evidence for VR, and “inconsistent evidence” for robotic training. Due to the lack of studies, understanding the effects of LL wearable technologies remains “very limited.” With newer technologies like soft wearable robotics, research in this area will grow exponentially. Future research can focus on identifying components of LL rehabilitation that can be effectively addressed using wearable technologies.

quality of life. For example, 80 percent of stroke survivors face motor impairments affecting one side of their body [1]. Previous research [2] indicated that only 15 percent of stroke survivors achieve full functional recovery in both limbs. In comparison, 33 to 60 percent have significant residual impairments in their hemiplegic arm at the chronic phase [3]. Though there have been recent advances in the medical management of stroke, most post-stroke recovery relies heavily on rehabilitation interventions [4].
Intensive rehabilitation has been shown to enhance motor recovery after stroke [5], and rehabilitation is necessary until maximum recovery is achieved [6]. Nevertheless, such extensive training is not sustainable in the long run due to the high cost of post-stroke care, such as rehabilitation services (i.e., therapist salary, rehabilitation site) and hospitalization [7]. Therefore, the self-management paradigm has been adopted to facilitate home-based self-directed training to reduce the burden placed on existing healthcare resources [8]. Selfdirected rehabilitation is conducted independently by patients and carers without direct supervision from a healthcare professional. This sort of training at home offers several advantages, such as providing contextual learning in real-life environments that promotes generalization [9], [10] and reduces the cost of supervised therapy [11].
The use of wearable technology is a promising option for providing home-based self-directed rehabilitation while keeping costs low [7]. Using wearable devices offers several advantages over conventional approaches. For example, some devices are portable, low-cost, and flexible [12], [13]. Wearable technologies are electronic hand-free devices worn externally on the body and monitor activities without limiting users' movements [14], [15]. In rehabilitation, wearable technologies are applied to measure body kinematics outside the laboratory and augment posture and motion correction by providing real-time feedback to users [13], [16] or assistance (passive or active assisted) in movements [17].
Some wearable devices provide real-time augmented feedbacks by emitting auditory, visual or tactile cues to the user, which is critical for motor relearning [16] and sustains motivation during training [18]. This feedback increases the awareness of correct posture and movement patterns during task execution [13], [16] in stroke individuals whose intrinsic feedback mechanisms (e.g., proprioceptive cues) are weakened or impaired [18]. Traditionally, the therapist provides extrinsic feedback to facilitate motor relearning in persons This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ with stroke [13]. However, this training method is very timeconsuming and manpower-intensive to carry out at home [13]. Alternatively, these wearable devices initiate augmented feedback to prompt individuals to perform self-directed training in the home setting. Unlike traditional methods of monitoring therapy adherence such as using an activity logbook or checklist, the wearable device increases treatment adherence in the home setting by providing objective feedback on the type and amount of upper limb training and trigger sensory reminders to increase the frequency of upper limb practice [19].
Increasing publication trends on the use of wearable technologies in stroke rehabilitation highlight the growing interest in this area [19], [20]. Maceira-Elvira et al. [7] and Kim et al. [19] conducted scoping reviews on using UL wearable sensors for assessment and treatment in the stroke population. They found that several studies had focused on hemiparetic UL measurement with sensors, but few focused on treatment approaches, and there is a lack of large-scale studies to prove the clinical efficacy of wearable sensors for home use [19]. Another study by Peter et al. [20] focused on reviewing the evidence of wearable sensors for gait assessment and did not look at other types of wearable devices or treatment uses. All these studies [7], [19], [20] narrowed their scope to wearable sensors; other wearable devices, such as stimulators and robotics, were not explored. Finally, two systematic reviews by Parker et al. [14] and Powell et al. [21] investigated the evidence of wearable devices for upper and lower limb rehabilitation, respectively. These reviews included other wearable devices for poststroke rehabilitation, such as electrical stimulation and robotics. Both studies [14], [21] revealed a paucity of high-quality evidence supporting using upper and lower-limb wearable technologies to improve activity and participation. Nevertheless, both narrowed their scope to select randomized controlled studies that used wearable devices to improve activity and participation. Other study designs and outcomes, such as motor impairment and function, were not addressed.
The effects of wearable devices in home-based stroke rehabilitation remained unclear from the analysis of previous reviews [14], [19], [20], [21], as most focus on wearable sensors, which are predominantly used for assessment rather than treatment [7], [19], [20]. Augmented feedback from a wearable device may make it an effective tool in motor training for stroke survivors beyond its measurement capabilities. In addition, the current evidence from previous reviews seems to skew toward using wearable technologies in care institutions or laboratories requiring the supervision of a rehabilitation specialist [15], [20], which eliminates stroke survivors' ability to self-direct their training [20].
Although earlier studies [7], [14], [19], [20], [21] contributed valuable knowledge to wearable technologies research, a comprehensive review of their application in home-based stroke rehabilitation remains scarce. To the best of our knowledge, no review has investigated the effectiveness of wearable devices as a treatment option in home-based rehabilitation for persons with stroke. A scoping review method is commonly used for new research areas because emerging and diverse evidence clarifies key concepts and characteristics and identifies research gaps [22]. This scoping review aimed to (1) map interventions that use wearable technologies in home-based physical rehabilitation for stroke, and (2) provide a synthesis of their effectiveness as treatment options. The findings of this review shed light on the research gap and aid researchers and clinicians by providing valuable knowledge to translate the use of wearable technologies into clinical practice.

A. Design
This scoping review adopted the framework outlined by Arksey and O'Malley [23], which involves identifying research questions and relevant studies, study selection, charting data, collating, summarizing and reporting results. The Preferred Reporting Items for Systematic Review Extension for Scoping Reviews (PRISMA-ScR) checklist [24] was used to ensure this review's robustness. In addition, the population concept context (PCC) structure [25] was used to identify key elements to conceptualize the scoping review: stroke (Population), wearable technology (Concept), and home (Context).

B. Search Strategy
A systematic search was conducted on four databases: MEDLINE (via PubMed), CINAHL, Cochrane Library, and Web of Science from inception to February 2022. In addition, the reference lists of eligible studies were hand-searched to identify any potential studies not identified through the database search. The searches were restricted to human studies. Search terms were "wearable," "rehabilitation," and "stroke" and their variations (please refer to Supplementary information, S1 for the complete list of search terms).

C. Selection Process
The selection process was followed according to the PRISMA guidelines [26]. The studies were selected based on the following eligibility criteria: Population: adults (> 18 years old) with stroke; Concept: In this review, wearable devices are defined as electronic "hand-free devices worn externally on the body that are portable (not fixed to a station, i.e., end effector) and can be used independently of a therapist" [14], [15] and studies must use wearable devices for treatment purposes; and Context: devices used in the home setting. Duplicates, didactic papers, posters, book chapters, study protocols, conference proceedings, systematic reviews, and meta-analyses were excluded. Studies that did not use stroke subjects and those not published in English were also excluded.
This review recognized that the development of a wearable technology system undergoes different levels of technological "maturity" and adopted the definitions given by Moral-Munoz et al. [27], see Supplementary information, S2. This review focused on studies that had already piloted a device in the home setting with persons with stroke. Hence, it included studies that coincided with level 5 in technological "maturity", After completing the database search, duplicates were removed by one reviewer (ST) using Endnote X9 [28]. Two independent reviewers (ST, PC) screened for study eligibility based on titles and abstracts retrieved during the searches using selection criteria. The two reviewers independently reviewed the full text of pre-selected articles and agreed on the final set of included studies. If there was a disagreement regarding the studies to be included, a third reviewer (KNKF) was consulted to achieve consensus.

D. Data Extraction and Quality Assessment
One reviewer (ST) extracted data from the selected studies using a data charting table, and the second reviewer checked the accuracy (PC). The extracted data included: author name, year, study design, sample size, participants, types of interventions, types of feedback, outcome measures, and results. The methodological quality of the selected studies was assessed using the National Institutes of Health (NIH) risk of bias tools for "controlled trial" and "before-after (pre-post studies) studies with no control group" [29] by two reviewers (ST and PC). Due to the limited generalizability of the nature of the case study, a case study design is considered a low-quality study in this review. After assessing the methodological quality of the included studies, synthesis was performed to evaluate their evidence level based on the following hierarchical criteria shown in Supplementary S2, as previously described in other studies [16], [30], [31], [32].
The total number of participants in the included studies was 717, with sample sizes ranging from 1 to 95. The mean age of the participants was 60.4 (5.3) years. Two studies [47], [55] did not report the age of their participants. The mean stroke onset of the participants was 34.8 (22.6) months. Seven studies [36], [43], [44], [46], [47], [55], [58] did not indicate the time of their participants' stroke onset.
1) Virtual reality: VR is a form of training where patients interact with a virtual or augmented environment created with the aid of technology [61]. There are two types of VR: immersive and non-immersive. Immersive VR rehabilitation provides a training environment that refocuses users' sensations from the real to the virtual world [62]. This type of training usually includes using a head-mounted display or goggles, which can screen out other stimulation from the virtual environment [63].  I  CHARACTERISTICS OF SELECTED STUDIES  TABLE II  RISK-OF-BIAS EVALUATION FOR RANDOMIZED CONTROLLED TRIALS (N=16) USING THE NATIONAL HEALTH INSTITUTES OF HEALTH RISK-BIAS  TOOL FOR CONTROLLED TRIALS   TABLE III  Unlike immersive VR, non-immersive VR allows users to see the screen environment [63] and interact with the VR task on-screen [64]. Examples of non-immersive VR rehabilitation are offered by Kinect, Nintendo Wii, and IREX. Only studies that used wearable devices in VR training were included in this review. Studies that used video-captured VR (i.e., Kinect, Nintendo Wii) without any wearable interface were excluded.
Nine of the included studies [34], [42], [43], [44], [47], [53], [54], [55], [56] used wearable sensors in VR training, all non-immersive. Nearly half of these studies [42], [43], [55], [56]) used information communication technology to monitor participants' progress remotely. The wearable sensors in the VR systems upload data to an encrypted cloud server, allowing the therapist to monitor the participant's progress and provide timely feedback remotely. Interactive therapy (i.e., VR) and remote monitoring promote engagement and sustain users' motivation, which is essential for unsupervised therapy at home. Cramer et al. [65] stressed that sustaining the participant's motivation in unsupervised home-based therapies is challenging, and previous studies have reported high nonadherence levels of up to 70 percent [66]. Remote monitoring from the therapist mitigates this risk of non-adherence by promoting therapist-client interaction and offering timely feedbacks [67]. Finally, all the VR studies used a monitor (computer screen or television) or tablet as their visual display. The use of more portable devices such as smartphones had yet to be explored.
2) Activity trackers: Seven studies [36], [39], [40], [51], [52], [57], [58] used wearable devices as activity trackers. Five [36], [39], [40], [51], [52] used these devices to track the use of impaired arms, while two (57,58) used them to track participants' physical activity. All the wearable devices used in these studies, except for one [57], resembled a wristwatch, and participants were instructed to wear it for a predetermined period. The study by Paul et al. [57] used a smartphone as a wearable device and instructed participants to use a pouch to carry with it. All the studies used the tracked activity data to promote participants' awareness of their activity level. The activity trackers in more than two-thirds of these studies [39], [40], [51], [57], [58] emitted visual and/or vibration signals to prompt participants to move their impaired arms or increase their physical activity level. Two studies [36], [52] used the tracked activity data to provide summary feedback to participants on their performance during the therapy visit.
3) Sensorimotor stimulation: Five studies [38], [41], [48], [49], [50] used wearable devices to provide sensorimotor stimulation to treat the hemiparetic arm. One study [59] used a full-body wearable suit to stimulate the upper and lower limbs. These studies used two types of stimulation: electrical and vibratory. Two-thirds of the studies (n = 4) [38], [41], [50], [59] used electrical stimulation on muscles or nerves, while the remaining two [48], [49] used vibration to stimulate the skin underneath the device (i.e., the dorsum of the wrist or hand). Unlike electrical stimulation, vibratory stimulation can be applied mechanically with or without the placement of an electrode. The two types of stimulation also differ in mechanism: electrical stimulation targets the tissue responses from the nerves or muscles [68], while vibratory stimulation targets the cutaneous mechanoreceptors underneath the skin and afferents [69], [70]. 4) Robotics: Five studies [33], [35], [36], [45], [46] applied wearable technologies in robotic training to improve hemiparetic UL function. Robotic devices are wearable interactive motorized devices that allow fine-graded limb movements [71] and provide passive, active assisted, or resistive training [17]. These wearable robotic devices enable repetitive, intensive, and task-specific training to promote motor learning [72]. Wearable soft robotics (WSR) is increasingly attracting researchers' interest in wearable technologies research [73]. WSR devices use soft and flexible garment-like materials, making them more lightweight [46] and offering more flexibility and versatility for the user's comfort and ease of use [73]. Nevertheless, this review found that the clinical application of WSR in home-based stroke rehabilitation remains limited. Due to its novelty, only one study [46] identified in this review used a wearable soft robotic glove to provide UL training for persons with hand limitations. Four other robotics studies [33], [35], [37], [45] used a rigid wearable hand exoskeleton called a handspring-operated movement enhancer (HANDSOME) or a mechatronic device called SCRIPT Prototype 1. This type of wearable exoskeleton resembles a wrist, hand, and finger orthosis with a passive actuation mechanism to aid users in grasping and releasing objects. In addition, three robotic studies [33], [45], [46] integrated their robotic devices into a VR system to increase the engagement of their participants.

D. Effects of Interventions Using Wearable Technologies
This review analyzed the quality of evidence on the effect of the four modalities mentioned above. The synthesis of the quality of evidence ( Figure 2 & Table I) identified a wide variation in the level of evidence concerning their effect in the different targeted regions, such as the hemiparetic UL, LL, and physical function.
Five RCT studies [38], [41], [48], [49], [50] contributed to the "strong" evidence of the effectiveness of sensorimotor stimulation in improving hemiparetic UL function. Among these studies, three [41], [49], [50] were rated as being of "high" quality, and two [38], [48] were of "fair" quality for their methodology. All the studies but one [49] reported improvements in the motor performance of the hemiparetic UL in their participants after stimulation. Seo et al. [49] did not assess their participants' arm performance. Instead, the authors [49] evaluated the safety aspect of applying a wearable vibrator at home, demonstrating the safe application of the device for long-term daily use at home. "Moderate" evidence was found in five studies [36], [39], [40], [51], [52] that used wearable technologies as activity trackers. Though all the studies supported the effectiveness of the activity trackers in improving impaired arm function, they varied in their study quality and design. There was one "high"-quality RCT [51], one "fair"-quality RCT [39], one "low"-quality non-RCT [36], and two "low"-quality "before-after" studies [40], [52].
2) Intervention effect on hemiparetic LL and physical function: This review revealed a paucity of studies that applied wearable technologies in home-based LL rehabilitation for persons with stroke. Only one case study [55] was found that applied a home-based telerehabilitation system supported by wearable insole sensors to improve the gait of a single subject. Though this study reported favorable outcomes, its level of evidence was regarded as "low." Therefore, the overall evidence in this area was considered "very limited". Similarly, "very limited" evidence from three studies [56], [57], [58] was available to support the effectiveness of wearable technology interventions in improving the physical function of stroke participants. Two studies [57], [58], a case study and a non-randomized controlled trial, used activity trackers to improve participants' physical function. Another "beforeafter" study [56] used VR training instead. Although all these studies [56], [57], [58] reported favorable outcomes for the participants' physical function, the overall evidence was considered "low"-quality due to the high risk of bias linked to the methods used.

IV. DISCUSSION
This scoping review mapped the current home-based stroke rehabilitation interventions that use wearable technologies and provided a synthesis of evidence concerning the effectiveness of these interventions. Four types of interventions that applied wearable technologies in home-based stroke rehabilitation were identified -VR, stimulation-based training, robotic therapy, and activity trackers. This review uncovered varying evidence concerning the effectiveness of these interventions in the different targeted regions, such as the hemiparetic upper limb, lower limb, and overall physical function. Most studies on wearable technology research in the home focused on improving the hemiparetic UL, while few concentrated on the lower limb region and overall physical function. In the following section, we discussed the effects of wearable technology interventions and future research directions.

A. Effects of Wearable Technology Interventions
Overall, most of the reported outcomes of interventions that used wearable technologies in stroke rehabilitation were positive, such as improved paretic UL function and increased walking and physical activity. Nevertheless, a synthesis of the evidence on the intervention effect on the hemiparetic arm highlights two key findings. This review found "strong" evidence supporting the effectiveness of stimulation-based interventions and an "inconsistent" level concerning wearable robotic training.
Somatosensory input provided by wearable stimulation devices appeared to be effective in improving hemiparetic arm outcomes. The effectiveness of somatosensory stimulation to improve arm motor function has been debated in previous systematic reviews that reported varying success [74], [75]. A meta-analysis [74] demonstrated that electrical stimulation effectively improved the hemiplegic arm function of persons with stroke. This study included the results of 48 RCTs and found that the electrical stimulation group showed more significant improvement in the affected arm than the control group. In contrast, another systematic review [75] reported low to moderate-quality evidence from 15 studies suggesting that somatosensory stimulation did not improve hemiparetic arm function. Nonetheless, it is notable that the review included studies that used thermal and compression therapy that shared different mechanisms from the electrical or vibratory stimulation used in the studies selected here. In this review, the selected sensorimotor stimulation studies used a high treatment dose (i.e., average total treatment dose: 102.5 hours). High doses provide enhanced somatosensory stimulation, which is believed to have a priming effect inducing changes in motor cortical excitability [75], [76]. Previous studies supported that enhanced somatosensory input from electrical stimulation improves the hemiplegic UL function in persons with stroke [41], [77]. Another recent RCT also found that vibration stimulation enhanced neural communication in the cortical sensorimotor network in their participants [78].
Another explanation is that sensorimotor stimulation such as electrical stimulation has been used as rehabilitation technology with hemiplegic patients for more than 30 years [71]. This technology is considered relatively mature and has been proved to be effective in stroke rehabilitation, where its clinical validity and safety are established. With a high level of technology maturity, more large-scale and well-designed studies are conducted, contributing to the high-quality evidence found in this review.
Another key finding is the "inconsistent" evidence concerning the effectiveness of wearable robotic training in home-based UL rehabilitation. This finding is consistent with previous reviews [79], [80] that showed the benefits of the effectiveness of robotic training over conventional therapy were debatable. The Cochrane review by Pollock et al. [80] and the review by Maciejasz et al. [79] found insufficient evidence supporting the effectiveness of robotic training over conventional therapy to improve the hemiplegic UL function.
Furthermore, applying robotic therapy in the home poses additional challenges to developers and researchers. Traditionally, robotic training has been used in the laboratory or hospital due to the cost and size of the equipment and the need for a skilled operator [81], [82]. One also needs to consider the variety of exercises that robotic therapy can offer to sustain participants' engagement. For instance, one RCT [45], the main contributor to the "inconsistent" evidence in robotic therapy, found a significantly lower adherence in the robotic group due to the limited variety of exercises offered by the robotic system. It has also been observed that fewer than half of the wearable robotics studies [45], [46] used an RCT design. This observation implies the existence of challenges in designing and developing a robotic system suitable for home use. Home-based robotic technology is still maturing, and further fine-tuning is needed.
This review revealed a paucity of research on wearable technologies in home-based LL rehabilitation. The literature search found that studies [83], [84], [85] that applied wearable technologies in LL stroke rehabilitation had predominantly carried out interventions in the laboratory or hospital setting. In addition, all these studies [83], [84], [85] were at the "proof of concept" stage with small-scale study designs such as case studies or single groups. The application of wearable technologies to improve LL outcomes seems to be at an early stage, such as level 2 of technology "maturity" described by Moral-Munoz et al. [27].
One intention of home-based technology-assisted interventions such as wearable technology is to offer a platform for unsupervised or under-supervised therapy, reducing the need for the physical presence of a therapist. This approach may not be suitable for certain types of LL rehabilitation, such as balance training. A previous systematic review [86] highlighted that remote supervision delivered through telerehabilitation was ineffective compared to usual care for balance training. Patients require adequate safeguards and physical assistance during balance training, so in-person rehabilitation may be more appropriate than remote mode [86]. Further studies can consider identifying the components of LL rehabilitation that can be effectively addressed by wearable technologies and those that require supervised therapy.

B. Future Research Directions
Smartphones' popularity and high ownership rate globally [27] offer new opportunities for VR rehabilitation. Smartphones are portable and offer more flexibility for users to conduct training in their preferred location. All the VR training in the selected studies used a monitor or tablet as the visual display, which requires some setup and takes up physical space in the home. Future studies can explore using smartphones as the visual display for VR training.
Stroke survivors learn to compensate for their motor impairment by not involving or using their hemiparetic UL in daily activities [87]; this is commonly known as a learned non-use phenomenon. The consequence is that therapeutic gains from the prescribed intervention deteriorate rapidly over time [87], [88]. One valuable feature of wearable sensors is that they can track activity and prompt individuals to use the impaired arm in their daily routine outside the "prescribed therapy" time. Furthermore, a similar type of sensor is used in VR training. Future development can consider harnessing the full potential of wearable sensors to develop a device that can support "prescribed" therapy in real environments and be used as an activity tracker to prompt users to use their impaired limbs outside therapy time. In this way, it addresses the need for intensive practice of the impaired limbs in daily routine to sustain therapeutic gains.

C. Limitations
This review was limited to the studies that used wearable technologies to provide stroke rehabilitation in the home. Furthermore, it focused on studies that conducted clinical trials on the target population (i.e., persons with stroke) with technological "maturity" of level 5. Other studies that have applied wearable technologies of a lower technology "maturity" level in the clinical or laboratory setting exist, but they were not within the scope of this review.

V. CONCLUSION
Wearable technologies have the potential to provide intensive home-based therapy in a self-directed manner for persons with stroke. With newer technologies such as soft wearable robotics and telerehabilitation emerging, research in wearable technologies for home use will grow exponentially. This review identified that most current research focuses on the UL, and there is a paucity of studies concerning LL rehabilitation. Future studies can consider identifying components of LL rehabilitation that can be effectively addressed by wearable technologies and those that require supervised therapy.
LIST OF ABBREVIATIONS PCC: population concept context; NIH: national institutes of health; RCT: randomized controlled trial; VR: virtual reality; UL: upper limb; LL: Lower limb; WSR: wearable soft robotic