Abstract
Rehabilitation devices are technologies that can automate repetitive features of therapy using force generation elements like motors to render training forces or gamified environments to improve user engagement (e.g., rehabilitation robots). These devices have received considerable attention from researchers and clinicians over the past several decades as a means to increase dosage of intensive rehabilitation following a stroke. However, the commercial results of these efforts often manifest as highly motorized, expensive, and bulky devices that are unsuitable for the majority of clinical or home environments. Indeed, as access to rehabilitation resources begins to reveal itself as a critical obstacle to recovery for many stroke survivors, it is important for researchers to examine alternative approaches to facilitate device adoption. A handful of researchers have attempted to bridge this gap with increasing success by designing affordable and portable devices for post-stroke rehabilitation. However, the methods employed to lower device cost are quite varied; therefore, a synthesis of these approaches could benefit other researchers. In this review, we discussed the field of rehabilitation robots and provided a review of 37 existing low-cost devices for stroke rehabilitation. These devices engage patients using a variety of actuation methods to produce training forces: Active (controllable actuator that adds or dissipates energy e.g., motors, stimulators), Passive (uncontrollable actuator that only dissipates energy e.g., springs, cables), Semi-Passive (controllable actuator that only dissipates energy e.g., brakes) and Augmented Feedback (no actuator). Following this review, we outline certain unexplored areas of low-cost devices that may be fruitful areas of future exploration.
Keywords
Introduction
Millions of individuals around the world suffer from a stroke every year (Global Burden of Disease Stroke Collaborators, 2021). Many stroke survivors suffer from motor impairments on one side of their body that can limit their ability to (i) incorporate their more-impaired upper extremity into daily activities and/or (ii) ambulate independently (Langhorne et al., 2009; Winstein et al., 2016). Neuroplasticity—the brain’s ability to reorganize the function of residual tissue in response to injury—has been increasingly recognized as the driving factor behind recovery of motor function following a stroke (Cramer, 2008). However, spontaneous neuroplasticity is often maladaptive, resulting in undesirable functional outcomes including learned disuse and interhemispheric inhibition (Lang et al., 2021; Murase et al., 2004). Behavioral interventions that facilitate experience-dependent neural plasticity are therefore necessary to guide neural reorganization towards more beneficial outcomes. The principles of experience-dependent neuroplasticity have been outlined previously (Kleim and Jones, 2008), and different training types (e.g., task-specific training, strength training) leverage these principles to varying extents. However, three principles are critical in all forms of rehabilitation: repetition, intensity, and time (Lang et al., 2015; Lohse et al., 2014; Waddell et al., 2014).
Unfortunately, the cost of outpatient therapy can inhibit stroke survivors’ access to a sufficient dosage of intense, repetitive training, and as a result, many stroke survivors do not achieve their full recovery potential. Indeed, lower-income individuals have a higher likelihood of stroke incidence and increased stroke severity (costs scale with severity) (Bray et al., 2018; Fan et al., 2023; Ghoneem, 2022; Kerr et al., 2011; Marshall et al., 2015). In the United States, the average annual cost of outpatient rehabilitation is over $16,000 USD ((Godwin et al., 2011), based on Medicare reimbursement rates and scaled for inflation from 2011 to equivalent value in 2024). Because of these high costs, many lower-income stroke survivors are unable to pay for large therapy dosages and must rely on insurance providers (e.g., Medicare, Medicaid, or private insurance) to cover the expenses. However, Medicare only initially covers around $5,000 in combined physical, occupational, and speech therapy (American Physical Therapy Association); therefore, many stroke survivors with insufficient coverage must either pay the difference out-of-pocket or opt to receive less therapy. Additionally, physical and occupational therapists in typical outpatient rehabilitation settings work with patients in one-on-one sessions; therefore, the cost of rehabilitation will increase as the population ages and stroke prevalence increases.
Rehabilitation Devices Aim to Augment Outcomes by Increasing Dosage
Rehabilitation devices—technologies that automate and augment features of stroke rehabilitation—may be able to bridge this gap between necessary and provided care to stroke survivors. Rehabilitation devices can be either free-standing or wearable that use active (e.g., motors) or passive (e.g., masses, elastic bands) elements to generate training forces as well as sensors (e.g., rotary encoders, force transducers, etc.) to detect and react to a patient’s movement. By automating and augmenting therapy, rehabilitation devices has the potential to dramatically increase a patient’s dosage of intensive training. Rehabilitation devices can also relieve some of the time commitment and physical burden of therapy on clinicians, who not only directly oversee day-to-day therapy of each of their patients but often manually assist/resist patient movements to augment therapy (Babaiasl et al., 2016). Additionally, computer-controlled rehabilitation devices equipped with sensors can easily monitor patient progress during recovery and integrate gaming interfaces to increase patient engagement and encourage more intense training (Augenstein et al., 2024; Babaiasl et al., 2016). These devices can also expand the possibilities of in-home therapy, and thus, dramatically grow the possible dosage of training.
However, many commercially available stroke rehabilitation devices are quite expensive due to their level of functionality and space/infrastructure requirements. For instance, the ArmeoPower (Hocoma AG, Rockland, MA, USA) is a popular upper extremity rehabilitation robot that costs over $100,000 (Aprile et al., 2019). ArmeoPower is a free-standing device with a robotic arm that acts as an exoskeleton, with attached cuffs for the user’s upper arm and forearm, as well as a handle. During operation, the user sits in a chair, and the robot provides separate motorized torques to the shoulder, elbow, and wrist that can be used for passive stretching, assisting motion (e.g., anti-gravity support), or resisting motion (e.g., error amplification) (Marchal-Crespo and Reinkensmeyer, 2009). The Lokomat (Hocoma AG, Rockland, MA, USA) is a popular, commercially available lower extremity rehabilitation device that costs more than $400,000 (Esquenazi, 2018). The Lokomat is a free-standing structure that features an integrated treadmill, body-weight support system, and exoskeletal components that provide bilateral assistance to the hip and knee joints. The conjunction of these components allows for similar controllers as the ArmeoPower applied to the lower extremity. In both cases, these devices attempt to provide maximum rehabilitation functionality, a goal reflected by their high cost. Indeed, a large portion of commercially available rehabilitation devices/robots are over $40,000 (Aprile et al., 2019).
The result of these high prices is a collection of commercially available devices that are not always aligned with end-user needs (Alqahtani et al., 2021; Manz et al., 2022; Wolff et al., 2014). As a result, clinics and patients are often unable to afford these devices and rely on passive equipment (e.g., exercise machines, towel-slides) for rehabilitation. Despite the effectiveness of this equipment, they do not automatically target stroke-related deficits and therefore require direct oversight from a therapist to monitor activity and track progress, thereby limiting the extent to which they can expand rehabilitation dosage.
Researchers and engineers can lower the cost and size of rehabilitation devices by reducing the extent of the device functionality. For instance, the Lokomat offers motorized assistance or resistance to both lower extremities as well as body-weight support, which represents the near maximum functionality of a lower extremity rehabilitation robot. Previous researchers have been able to reduce the size and cost of rehabilitation devices relative to the Lokomat by only providing body-weight support (Frey et al., 2006), targeting only the more-impaired limb (Hidayah et al., 2020), or by targeting only the hip (Lee et al., 2020), knee (Puyuelo-Quintana et al., 2020; Sulzer et al., 2009), or ankle (Swaminathan et al., 2023) of the more-impaired limb. Similarly, previous researchers have reduced the size and cost of upper extremity devices as compared to ArmeoPower by targeting only the shoulder (Trigili et al., 2019), elbow (Vitiello et al., 2012), wrist (Maeda et al., 2024), or hand (Brokaw et al., 2011; Farrell et al., 2007; Lambercy et al., 2011) of the more-impaired limb.
Rehabilitation device cost can also be reduced by interacting with the patient at their end-effector rather than applying separate torques at each joint (Washabaugh and Krishnan, 2022). For instance, upper extremity devices that assist/resist the motion of the wrist joint center can target deficits at the shoulder and elbow (Campolo et al., 2014; Mazzoleni et al., 2018; Miao et al., 2017; Rosati et al., 2007). Likewise, a device that applies a force to the ankle joint center can target the hip and the knee simultaneously (Washabaugh and Krishnan, 2022). By targeting multiple joints with a single force, these devices can reduce the number of necessary actuators, and thus, the cost of the device. Furthermore, exoskeletal devices that apply torques to each of the user’s joints typically necessitate length adjustment mechanisms that can increase cost and reduce accessibility. An end-effector device, on the other hand, can be free-standing or table-mounted, thereby removing these adjustment mechanisms and allowing for easier access. Wearable devices are also vulnerable to soft tissue movement and imperfect adjustment that can lead to energy losses (Washabaugh et al., 2023), necessitating more powerful actuators to achieve desired torques. End-effector devices, on the contrary, generally attach to a bony area like the wrist, hand, or ankle and therefore are more efficient in energy transfer. However, end-effector devices do compromise the ability to strategically target each joint, so careful considerations in force selection are necessary to strategically target stroke-related deficits (Augenstein and Krishnan, 2022).
Despite these advancements by researchers to reduce the cost and size of rehabilitation technologies, most of these devices are still too expensive for most clinical or in-home environments. Indeed, an international survey of 233 physio- and occupational therapists from Australia, Canada, and the USA found that clinicians generally believed that a rehabilitation device such as a robot should cost no more than $6,000 (Lu et al., 2011). Thus, one promising approach to expanding rehabilitation dosage is by creating compact, low-cost rehabilitation devices. Such devices would be more accessible to a clinic and more likely to make it into a person’s home for daily use. However, designing such a device requires careful consideration on the part of researchers to reduce cost and size as well as minimize the impact of these reductions on the function of the device and its utility in rehabilitation. In this article, we will review tactics researchers and engineers employ to reduce the cost of these devices.
Materials and Methods
Search Strategy
To generate the initial article pool for screening, we systematically searched and manually retrieved articles from MEDLINE (Medical Literature Analysis and Retrieval System Online, accessed via PubMed), EMBASE (Excerpt Medical Database), Cochrane Central Register of Controlled Trials (CENTRAL), and Web of Science published between January 1st, 1990 and June 1st, 2024. The search terms used keywords from five conceptual domains: low-cost (e.g., inexpensive, affordable), device (e.g., robot, technology), stroke (e.g., hemiparesis, CVA), rehabilitation (e.g., physiotherapy, occupational therapy), and outcome measures (e.g., electromyography, kinematics). The full list of search terms and Boolean combinations is available in the Supplemental Material. The literature search and screening procedures were informed by a separate, ongoing, previously registered meta-analysis (PROSPERO Record: CRD42024554708) that quantitatively examines the effectiveness of low-cost robotic rehabilitation devices in post-stroke rehabilitation. The current review aims to synthesize the engineering strategies used to develop low-cost devices, and thus, has a distinct objective.
Eligibility Criteria
Article eligibility was determined based on the type of study, participant population, sample size, rehabilitation device description, and measured outcomes. Regarding the type of study, we limited our analysis to original randomized, controlled trials (parallel- or crossover-designed) and single-group (uncontrolled) trials published in peer-reviewed journals. We excluded studies not written in English, case studies or series, observational studies, and conference proceedings. Following this, the study must have involved at least 6 individuals with non-cerebellar stroke interacting with a low-cost rehabilitation device designed to target post-stroke impairment or functional deficits in the upper or lower limb. For controlled studies, acceptable control conditions included standard rehabilitation care or using device in a control mode (e.g., device off). Studies with control groups that received no intervention were considered uncontrolled. Lastly, the included studies must have examined clinical measures of extremity impairment (e.g., Fugl-Meyer Scale) or functional level (e.g., Action Research Arm Test), muscle activity, kinematics, kinetics, neural activity (e.g., corticospinal excitability), or spatiotemporal gait variables.
Here, we considered a device to be low-cost if a primary motivation (i.e., active ingredient (Ranganathan et al., 2022)) of device development was to be less expensive than existing rehabilitation technologies. This motivation must have been explicitly stated by the authors of the study in the Title, Abstract, Keywords, Introduction, or Materials and Methods sections, in a related publication by the same authors, or in a publication by the inventors of the device under consideration. This approach to defining low-cost was used instead of a specific monetary amount to prevent our subjective opinions of what constitutes “low-cost” and availability of device costs from affecting analysis. Importantly, this criterion does not include devices whose authors/inventors state that their device is low-cost compared to usual care with a therapist.
Data Management and Study Selection
Search results were uploaded to Rayyan, which is a web service for systematic review management. We performed an initial screening of study Titles and Abstracts to eliminate studies with obvious violations of our inclusion criteria (e.g., systematic review, non-English language). We then determined the eligibility of the remaining articles using the Abstract, Introduction, and Materials and Methods sections. If an article was found during our review of the eligible manuscripts or similar reviews that fit our eligibility criteria, then it was also screened and assessed for eligibility.
Results
A flow chart of our literature search is shown in Figure 1. The initial search identified 801 articles that matched our search criteria. 297 of these articles were duplicates and thus removed. Following screening of the initial search results and additional potentially eligible manuscripts found during our review, 180 manuscripts were screened for eligibility. 47 manuscripts met our eligibility criteria and were thus included in this review. Because multiple studies examined similar off-the-shelf solutions, these 46 manuscripts contained 37 unique low-cost devices.
Flowchart of device identification.
Table 1 provides an overview of existing low-cost technologies for stroke rehabilitation included in this review. The studies with a check in the “RCT?” column were randomized controlled trials (using either a between-group or cross-over study design) that measured sustained changes in clinical measures of impairment, disability, and/or biomechanical/physiological variables from working with their device. All other devices were examined in either uncontrolled or non-randomized experiments. For studies that compared their device against a control condition, both superiority and equivalence were considered favorable outcomes. Equivalent outcomes were also considered as favorable because the aim of low-cost devices is to improve therapy dosage; therefore, the device could lead to improved outcomes in practice by allowing greater dosage.
Existing Low-Cost Devices for Stroke Rehabilitation in Literature.
RCT = Randomized Controlled Trial, * indicates that name is descriptor, not author-given. For “Solution Type”, “Novel Device” indicates that entire device is custom designed, “Software” indicates that the device is a software solution using existing hardware, and “Hardware” indicates that the device is composed of existing hardware refit for a new purpose.
It is important to note that many low-cost research prototypes and commercial devices were excluded from our review because their accompanying publications did not meet our inclusion criteria. Often, this occurred if the device was developed to be low-cost, but its effectiveness has not yet been examined in at least six stroke survivors. However, their engineering approach to reducing device cost may also be valid. Additionally, there were many devices that a reader may consider to be low-cost, but the authors/inventors do not identify the device as low-cost themselves. Thus, this review is not meant to be an exhaustive list of all low-cost rehabilitation technology, but rather to provide a succinct picture of the techniques/approaches that researchers and inventors use to produce low-cost devices that address post-stroke deficits.
Active Devices
A common type of low-cost rehabilitation device are those that provide external, supplemental power to the more-impaired limb. These devices belong to a broader category called active systems: systems composed of controllable force generation elements that can add power to or remove power from the patient–device system.
Motorized Devices
Electrical motors are the most common means to provide supplemental power in low-cost active devices. This is understandable, as electrical motors are readily available for purchase, more easily controlled than other actuator types (e.g., pneumatics), and offer a great deal of flexibility in device functionality. Supplemental power from an electrical motor can be useful for assisting individuals whose impairment interferes with/prevents functional usage of the limb. Motorized assistance can also generate joint movement with no user effort, which can be beneficial to improve joint, muscle, and tendon mobility and prevent joint pain in severely impaired individuals (Celik and Kaya Mutlu, 2016; Hesse et al., 2008; Kim et al., 2019; Longo et al., 2012). Additionally, devices with back-drivable motors can be easily reprogrammed to increase the difficulty of training by resisting patient motion. This form of resistance can be useful for stroke survivors with less severe upper extremity impairments to scale training intensity with recovery, increase engagement, and inhibit slacking (Reinkensmeyer et al., 2009). This resistance represents a form of functional resistance training: resistive loads are integrated into a functional context to leverage gains from both task- and resistance-oriented training.
In general, motorized devices reduce cost by limiting the number of motors in their system. This approach is not surprising, as motors generally increase cost due to additional equipment (e.g., power amplifiers) and safety considerations (Augenstein et al., 2024; Sung et al., 2017; Washabaugh and Krishnan, 2022). The low-cost motorized devices in Table 1 include the hCAAR (Sivan et al., 2014), the CR2-Haptic (Khor et al., 2017), the H-man (Budhota et al., 2021), the TheraDrive (part of the Robot Gym (Bustamante Valles et al., 2016; Johnson et al., 2004)), and the Haptic Cane Device (Afzal et al., 2018). The hCAAR, CR2-Haptic, H-man, and TheraDrive integrate one or two electrical motors that assist planar upper extremity end-effector motion, and all but the hCAAR offer the ability to resist motion. The devices that assist or resist motion typically provide assistance until it is apparent that the patient is not being challenged, after which the device resists their motion. The CR2-Haptic offers the ability to reconfigure the motion plane so that the user can receive assistance in different motions while maintaining a simple device. TheraDrive is a motorized force-feedback steering wheel that can be integrated into an interactive steering environment (Johnson et al., 2004). The only motorized low-cost device that provides assistance to the lower extremities was the Haptic Cane Device: a motorized cane that can either assist or resist a patient in maintaining a desired gait speed as well as provide kinesthetic cues to help regulate the amount of force a patient puts on their cane (Afzal et al., 2018; Pyo et al., 2015).
Evidence from studies examining the effectiveness of low-cost motorized devices for stroke rehabilitation indicates that devices with fewer motors can still lead to gains in clinical and kinematic measures. The hCAAR and CR2-Haptic have been examined in single-arm studies and were shown to improve clinical measures of impairment (e.g., UE-FM) and disability (e.g., Action Research Arm Test or ARAT) and kinematic measures following an intervention (Budhota et al., 2021; Khor et al., 2017; Sivan et al., 2014). The H-man was compared with dose-matched standard-of-care in a randomized controlled trial and was found to have similar improvements in clinical measures as the control group and an advantage in several kinematic measures (Budhota et al., 2021). The effectiveness of the TheraDrive in stroke rehabilitation has only been evaluated as part of a publicly available robot gym designed to reduce treatment cost (i.e., one device in a circuit of device) (Bustamante Valles et al., 2016); therefore, it is difficult to discern its unique effect in the intervention. The Haptic Cane Device has been shown to improve stance symmetry and increased activation of the quadriceps and hamstrings on the more-impaired limb (Afzal et al., 2018).
Stimulation-Based Devices
Some low-cost active devices stimulate the sensorimotor loop to facilitate recovery rather than providing training forces via electrical motors. The stimulations can include neuromuscular electrical stimulation (NMES) or peripheral sensory stimulation. When NMES is used to improve performance in a functional task, it is called functional electrical stimulation (FES).
Of the low-cost devices included in this review, only one involved FES to provide joint control assistance. This device, the Re-Lift, is an FES-based system that recognizes the user’s intent to perform a motion via a custom-developed tilt sensor and assists ankle dorsiflexion with a low-current electrical stimulation to the peroneal nerve (Sung et al., 2017), similar to commercially-available FES systems like the Bioness (Moody Neurorehabilitation Institute, 2024). FES systems are lighter and more compact than motorized systems and can therefore more easily assist lower extremity joints. Peroneal nerve FES systems have been previously shown to be as effective as ankle-foot orthoses (AFOs) in improving gait speed and measures of activity restriction after stroke (Bethoux et al., 2014). Additionally, electrical stimulation can reverse conversion of fatigue-resistant, Type I muscle fibers to fast-fatiguing, Type II muscle fibers that is characteristic of disuse muscle atrophy following a stroke (Kubis et al., 2002; Stein et al., 2014; Vanderthommen and Duchateau, 2007). A previous study with this low-cost FES system found that it improved ankle ROM during use compared to no device and performed similarly to a commercially available device (Ojha et al., 2023).
Additionally, there was one device that used low-cost peripheral sensory stimulation to facilitate recovery. This device, the TheraBracelet (Lakshminarayanan et al., 2017; Scronce et al., 2023; Seo et al., 2014; Seo et al., 2019), is a wearable wristband that provides a sub-sensory threshold random-frequency vibration to the wearer’s wrist during functional activities with the objective of increasing adherence to home-based therapy. Afferent stimulation has been previously shown to increase cortical excitability, and the sub-sensory random vibration provides this stimulation while not interfering with functional activities. The intensity of the signal is amplified in the afferent sensory pathway by stochastic facilitation, a phenomenon where certain frequencies of random noise are amplified in nonlinear systems (Collins et al., 1996; Scronce et al., 2023). A randomized controlled study with this device found that it led to greater reductions in disability as measured by the Box and Block test when compared with same intervention without the stimulation (Scronce et al., 2023).
Passive Devices
Because external sources of power like electrical motors often increase the cost of rehabilitation devices, many researchers opt for alternative methods to augment post-stroke therapy. These alternative methods typically result in passive devices: devices that use uncontrolled passive force generation elements (e.g., weights, elastic bands, cables) to augment training. Passive devices recycle or dissipate the energy supplied by the user to augment training but cannot add energy themselves. Additionally, because the force generation elements are uncontrolled, forces can be calibrated prior to use but cannot be adjusted in real-time.
Passive force elements can be particularly useful if the objective of the device is to resist a user’s motion to increase training difficulty. A device that can only resist motion may not be appropriate for severely impaired stroke survivors who require assistance to complete task-oriented training activities. However, for stroke survivors with mild-to-moderate levels of impairment who do not require motorized assistance, resistance can be a useful approach for progressing task difficulty. Therefore, eliminating motorized assistance reduces the cost without losing much functionality for these individuals.
Gravity Compensation Devices
Gravity compensation devices are the most common low-cost stroke rehabilitation device, whether passive or otherwise. These devices reduce the magnitude of joint torques necessary for functional usage of a limb. This allows stroke survivors with unilateral weakness or abnormal, torque-dependent joint coordination (e.g., the flexor synergy) to more easily integrate their more-impaired limb into task-oriented training, thereby increasing their success rate in performing goal-directed activities with normal kinematics. This increase in success rate is posited to facilitate skill learning in severely impaired individuals whose level of impairment/disability may prevent them from learning the skill without assistance (Marchal-Crespo and Reinkensmeyer, 2009). Following these theoretical motivations, the hope is that this assistance during training can be progressively reduced over a long-term intervention, thus translating to improvements in impairment, disability, and activity levels after the assistance has been removed (Marchal-Crespo and Reinkensmeyer, 2009).
Due to this theoretical motivation of improving limb performance in daily tasks, low-cost gravity compensation devices usually target the upper extremity. For the upper extremity, gravity compensation is often accomplished via a low-friction platform that supports the patient’s end-effector during movements in the transverse plane, similar to towel-slide exercises used during conventional rehabilitation. These devices include the RAPAEL Smart Board ((Park et al., 2019) commercialized by Neofect, Waltham, MA, USA), the ArmAssist (Tomic et al., 2017), and the AbleX Armskate ((Jordan et al., 2014), previously Smart Skate, commercialized by AbleX Healthcare, Auckland, NZ), although for many other devices covered in this review (Budhota et al., 2021; Bustamante Valles et al., 2016; Hesse et al., 2008; Khor et al., 2017; Ozawa et al., 2013; Sivan et al., 2014), this form of assistance is incidental (e.g., a motorized horizontal planar robotic arm will provide this assistance regardless of what the motors do). Patient movements with these devices are either unrestricted in the transverse plane or confined to configurable paths via placeable stoppers (Park et al., 2019), and are typically accompanied by biofeedback (Barker et al., 2008) or gaming interfaces/ADL-simulators (House et al., 2016b; Park et al., 2019; Tomic et al., 2017) to challenge the patient and enhance engagement.
Of these devices, only the RAPAEL Smart Board and the ArmAssist were examined in a randomized controlled manner with gravity compensation as the primary mode of patient interaction. Patients who practiced with ArmAssist were found to improve their upper extremity Fugl-Meyer (UE-FM) and Wolf Motor Function Test (WFMT) scores relative to the control group. The RAPAEL Smart Board was found to improve patients’ impairment and disability (measured with UE-FM, WFMT, and others) relative to baseline, but not more so than conventional occupational therapy.
Researchers have also developed more complex devices that provide gravity compensation to multiple joints simultaneously. These devices include the ArmeoBoom ((Prange et al., 2009) previously Freebal, commercialized by Hocoma AG, Volketswil, CH), a free-standing structure with suspended slings, the SpringWear (Chen and Lum, 2018), and the ArmeoSpring ((Housman et al., 2009) previously T-WREX, commercialized by Hocoma AG, Volketswil, CH). The latter two are wearable exoskeletons whose elastic torques assist in raising and extending the more-impaired limb. It is important to note that while these devices were designed to be low-cost as compared to existing rehabilitation devices, the cost of commercialized version of either device can exceed $40,000 USD, which some may not consider “low-cost” (Aprile et al., 2019).
These more complex gravity compensation devices have shown that the reduced gravitational loads can lead to improved task execution, which may facilitate recovery. When wearing the ArmeoBoom and SpringWear, stroke survivors showed online improvements in upper extremity kinematics with reduced muscle activation (Chen and Lum, 2018; Prange et al., 2009). Of these three devices, only the ArmeoSpring was evaluated with a randomized controlled study that found that it led to improvements in UE-FM, range of motion (ROM), and the components of the Rancho Functional Test as compared to a standard-of-care control group (Housman et al., 2009).
There was one low-cost device that provided gravity compensation during gait, AccesSportAmerica gait trainer, which involved refitting an exercise bicycle to provide gait assistance (Ventura et al., 2019). Although this device has been examined with stroke survivors, the data was pooled with data from participants with other neurological conditions and therefore cannot be discussed here. It should be noted that there are other low-cost devices for lower extremity gravity compensation that have been developed (Bu et al., 2023; MacLean and Ferris, 2020, 2021), although their clinical efficacy has not yet been examined.
Self-Power
Some low-cost devices incorporate a “self-power” paradigm to assist the more-impaired limb i.e., coupling the movement of the more- and less-impaired limbs so that the less-impaired limb assists the more-impaired limb. These systems reduce cost because the total power of the patient-device system does not change, therefore they avoid additional equipment and safety considerations inherent to their motorized counterparts. Self-powered devices in Table 1 include the eMBot (Washabaugh et al., 2018) and the ableX Handlebar ((Hale et al., 2012; Hijmans et al., 2011) previously CyWee Z, commercialized by AbleX Healthcare; Auckland, NZ). The eMBot couples elbow flexion of the more- and less-impaired limbs via an inverting transmission system composed of Bowden cables and pulleys. The ableX Handlebar, on the other hand, accomplish this with a bimanual handle that the patient grips with both hands. The ableX Handlebar is instrumented with a Nintendo Wii controller to track patient motion and integrate limb movements into gaming interfaces.
Another advantage of self-powered devices is that they do not add power to the patient-robot system; therefore, they are able to mitigate some of the detrimental effects of motor slacking by encouraging more active participation of the more impaired limb and potentially facilitate recovery (Washabaugh et al., 2018). A previous study with the eMBot has shown that self-powered training leads to increased muscle activation in the more-impaired limb as compared to externally powered training (Washabaugh et al., 2018). The ableX Handlebar was evaluated in a non-randomized repeated measures study wherein the control condition consisted of computer games played with the less-impaired limb, and ableX Handlebar was found to lead to greater improvements in upper extremity Fugl-Meyer scores (Hijmans et al., 2011).
Exosuits
Exosuits are a class of passive rehabilitation devices that involve systems of elastic bands/springs that span across the user’s joints; the elastic springs deform during a task and return the participant’s expended power as assistance at a later point. Although similar to the spring-based gravity compensation devices described in the previous section, there are a few key practical differences between those devices and exosuits. First, exosuits are typically used to assist the more-impaired lower extremity during gait, whereas gravity compensation devices focus on the upper extremity. Next, gravity compensation devices assist motion, whereas exosuits can assist or resist motion, depending on the set-point (equilibrium angle) of the elastic bands. For instance, if the ankle’s set-point is the neutral position, the exosuit would produce a restorative plantarflexion torque during dorsiflexion in late stance that could help propel the wearer and produce a restorative dorsiflexion torque during swing to prevent excessive plantarflexion (i.e., post-stroke foot drop) from interfering with gait. However, a high dorsiflexion set-point would assist dorsiflexion while resisting plantarflexion over the same gait region.
The exosuit included in this review, KickStart (Yao et al., 2021), incorporates elastic bands to assist/resist sagittal plane motion of the hip, knee, and ankle in both lower extremities, although similar low-cost devices exist for both the upper and lower extremities that have yet to be examined in stroke (Augenstein et al., 2025; AxioBionics, 2024; Krishnan et al., 2024; Maguire et al., 2010; Martins et al., 2019). A previous study with the KickStart has shown that it increased preferred gait speed, although these were online effects and no study has examined if it produces sustained effects (Yao et al., 2021). As noted above, exosuits could be easily deployed in stroke rehabilitation to provide resistance to joint motion during gait. This could be done by changing the set-points or providing visual feedback such that the user must stretch the elastic bands to achieve a desired gait cycle (Foley and Washabaugh, 2024; Washabaugh et al., 2020). However, this has yet to be examined in stroke using these devices.
Tone Compensation
A few low-cost devices exist that assist the user by targeting hypertonia following a stroke. These devices generally target the wrist and fingers of the more impaired limb and include the HandSOME II (Casas et al., 2021) and a passive stretching disc (Jang et al., 2016). The HandSOME II uses springs to assist finger extension so that a user with overactive finger flexors can more easily grasp a variety of household objects, similar to the commercially-available SaeboFlex (Barry et al., 2012; Farrell et al., 2007). The passive stretching disc, on the other hand, assists the wearer in wrist and finger stretching exercises prior to functional usage of the limb.
Tone compensation devices are generally effective at improving a user’s ability to incorporate their hand into functional activities. The online effects of the HandSOME II include increased finger extension and functional task success rate, albeit with reduced finger flexion. The sustained effects of the HandSOME II have been evaluated in a similar predecessor and found that UE-FM and ARAT improved after the intervention, although no control group was included in the study design (Brokaw et al., 2011; Chen et al., 2017). The passive stretching disc has been found to improve scores on UE-FM and the modified Ashworth scale (a measure of muscle tone), although the control group in this study received no intervention so it is unclear how this device compares to other interventions.
Inertial/Gravitational-Based Resistance
The above-mentioned passive device types aim to assist a user’s motion during functional use of the limb/limbs. However, as mentioned in the Active Devices subsection, many devices seek to generate forces during task-oriented training that resist their motion to increase training difficulty. Thus, training with these devices often represents a form of functional resistance training. As with the assistive devices, many researchers will eliminate motors as the primary force generating element from their devices and integrate alternative methods of resistance to lower cost. This is a particularly useful strategy for resistive devices because, as mentioned previously, stroke survivors with mild-to-moderate levels of impairment will often not require assistance to complete tasks. Therefore, eliminating motorized assistance reduces the cost without losing much functionality for these individuals.
Patient-based examinations of passive devices that add mass to the more-impaired limb have generally shown promising results. The NeuroBall is currently being examined in a clinical trial, although preliminary results indicate that an intervention with NeuroBall can lead to improvements in clinical measures of impairment, disability, and participation in daily life (Kilbride et al., 2022). In a randomized controlled trial, the SMART Arm improved stroke survivors’ clinical measures of disability, kinematics, and strength measures and showed greater outcome gains than the control group, although it is important to note that the control group did not receive any intervention.
Brakes
Two passive devices used brakes to increase resistance during training: a wearable leg brace (Washabaugh and Krishnan, 2018) and the Reha-Slide ((Hesse et al., 2008), commercialized by Deltason Medical Limited, Hong Kong, CN). The wearable leg brace uses an eddy-current brake design in which the knee’s joint motion during gait back-drives a planetary gearbox whose input is attached to the brake disc. By back-driving the gearbox, the knee joint velocity and brake torque are both amplified, thereby enabling the device to have a power density that is much greater than most similar motorized systems. The Reha-Slide, on the other hand, incorporated passive rubber brakes that could be enabled prior to training and provided 5-80 N of braking force during upper extremity reaching, although it should be noted that the Reha-Slide also features inclination and a bimanual handle for additional gravitational resistance and self-powered assistance, respectively.
Studies evaluating the efficacy of braking devices have shown that this resistance can lead to improvements in clinical, kinematic, and muscle activity measures. The randomized controlled study using the Reha-Slide showed improvements on clinical measures of impairment and disability similar to those gained from electrical stimulation therapy of the wrist extensors (Hesse et al., 2008). A study examining the wearable leg brace in chronic stroke survivors found that it led to increased gait speed, knee excursion, and hamstring activation after the device had been removed.
Semi-Passive Devices
While passive systems can circumvent many of the costly components associated with active systems, a noteworthy drawback is that the force generation elements they use are uncontrolled, meaning that device forces can be calibrated prior to used but cannot be altered in real-time. However, it may be possible to substitute these elements for controllable passive elements to create a unique class of rehabilitation robots: semi-passive devices. Semi-passive devices use controllable passive force generation elements (e.g., computer-controlled brakes, continuously variable transmissions, etc.) to control the level of assistance/resistance in real-time (Chang et al., 2018; Dellon and Matsuoka, 2008; Hirata et al., 2010; Swanson and Book, 2003; Takesue et al., 2018). A semi-passive device, therefore, offers the opportunity for unique control algorithms and training paradigms that a passive device cannot provide while still being lower cost than an active system. Critically, a semi-passive device may have motors included in the design (e.g., a motor may control the level of resistance provided by a brake); however, the motor output torque is not felt by the user/wearer.
Only one of the devices in Table 1, the PLEMO-P3 (Ozawa et al., 2013), is a semi-passive device, although a semi-passive version of the eddy-current leg brace has been developed but not yet examined in stroke survivors (Washabaugh et al., 2023; Yang et al., 2025). The PLEMO-P3 is a kinematic linkage that applies braking forces during reaching, although it differs from passive systems like the Reha-Slide because it can resist movement in any direction in its two-dimensional workspace, while the Reha-Slide can only resist motion parallel to its inbuilt linear track. To accomplish this, the PLEMO-P3 uses two controllable electrorheological (ER) brakes and a kinematically redundant structure to expand the range of possible end-effector velocities that it can resist (Furusho et al., 2002). This approach is similar to other passive and semi-passive force reflecting and haptic devices (Dellon and Matsuoka, 2008; Hirata et al., 2010; Swanson and Book, 2003; Takesue et al., 2018), although the PLEMO-P3 is the only one that has been used in stroke rehabilitation. Previous studies with PLEMO-P3 have shown that it can lead to improvements in UE-FM greater than standard of care (Ozawa et al., 2013). Semi-passive devices in stroke rehabilitation present a variety of unique rehabilitation paradigms that have previously only been done on active systems (e.g., guidance along a path, steering forces to alter muscle coordination), although to date only simple resistance has been examined.
Augmented Feedback Devices
Augmented feedback devices typically aim to increase the effectiveness of task-oriented practice for stroke survivors by increasing the patient’s engagement with practice. In this sense, they differ from active, passive, and semi-passive systems because they have no force generation element, although many of the systems mentioned above do incorporate some form of augmented feedback. Increasing engagement is typically accomplished in two ways: interactive, game/activity-based platforms that make repetitive training more exciting or biofeedback systems that encourage the user to practice certain desirable movement qualities.
Off-the-Shelf equipment
Devices that provide interactive, game/activity-based platforms using commercially available video gaming systems are the most common type of low-cost augmented feedback device. These devices use technologies such as the Nintendo Wii (da Silva Ribeiro et al., 2015; Hijmans et al., 2011; Mouawad et al., 2011; Neil et al., 2013; Warland et al., 2019), Microsoft Kinect (Colomer et al., 2016; Ikbali Afsar et al., 2018; Kim et al., 2018; Kizony et al., 2017; Thielbar et al., 2020), the Playstation EyeToy (Neil et al., 2013; Yavuzer et al., 2008), and immersive virtual reality headsets (NeuroVirt, (Mares et al., 2025) commercialized by NeuroVirt Limited, London, UK). These systems use the motion sensing capabilities of these commercially available systems as input for gaming interfaces or simulated ADLs that require usage of the more-impaired limb to succeed in the game.
Leveraging commercially available technology is useful for making a device low-cost as they provide high-end, robust motion detection and are sold at a competitive price, making them more accessible to patients, clinicians, and researchers. Several of these systems do not even make custom alterations to the device, and therapy simply involves using the games preloaded onto the device by the manufacturer (e.g., Wii Tennis (da Silva Ribeiro et al., 2015; Mouawad et al., 2011)). Furthermore, using commercially available technology increases the likelihood that a patient already owns this equipment and uses it regularly for its originally intended purpose (e.g., video games), decreasing the initial investment and learning time. Indeed, one Kinect system called the Virtual Environment for Rehabilitative Gaming Exercises (VERGE) connects multiple users from different physical locations to the same virtual environment to participate in collaborative/competitive activities to increase engagement (Thielbar et al., 2020; Triandafilou et al., 2018). However, one drawback of systems that use commercially available technology is that they are vulnerable to decisions of the manufacturer who may decide to discontinue the technology. Indeed, the Nintendo Wii, the Microsoft Kinect, and the Playstation Eyetoy are all no longer manufactured and can now only be acquired through third-party vendors.
Interventions with these systems have typically found that they improve clinical measures of impairment and disability as well as joint range of motion (Colomer et al., 2016; da Silva Ribeiro et al., 2015; Hijmans et al., 2011; Kim et al., 2018; Mouawad et al., 2011; Neil et al., 2013; Thielbar et al., 2020; Warland et al., 2019; Yavuzer et al., 2008). Two systems showed an advantage over conventional physical therapy in a randomized controlled trial (Ikbali Afsar et al., 2018; Yavuzer et al., 2008). In a randomized controlled trial comparing the multiple user experience with VERGE to a single user experience (i.e., same activities, but no additional connected users), the authors observed an advantage in more-impaired arm movement when connected with multiple users (Thielbar et al., 2020). To date, only NeuroVirt’s accessibility, usability, and activity level has been examined in stroke survivors (Mares et al., 2025).
Custom-Designed Equipment for Augmenting Therapy
Many game/activity-based platforms use custom-designed equipment to increase the number of possible activities for stroke survivors, and a large portion of these have been commercialized. These systems include the MusicGlove ((Friedman et al., 2014) commercialized by FlintRehab, Irvine, CA, USA), the SaeboReJoyce ((Ozen et al., 2021) commercialized by Saebo, Inc., Charlotte, NC, USA), the FitMi ((Swanson et al., 2023) commercialized by FlintRehab, Irvine, CA, USA), the Neofect Smart Glove ((Shin et al., 2022) previously RAPAEL Smart Glove, commercialized by Neofect, Waltham, MA, USA), and the GripAble ((Broderick et al., 2021) commercialized by Gripable Limited, London, UK). The MusicGlove and the Smart Glove are both sensorized gloves for the more-impaired hand that detect finger and hand movements and incorporate them into gaming interfaces. The Smart Glove additionally includes the ability to track progress across sessions to see improvement in function. The ReJoyce is a specialized manipulandum that includes attachments meant to emulate functional activities (e.g., a turnable knob, a key, etc.), and interaction with these attachments is meant to facilitate improvements in activities of daily living. The manipulandum is attached to a table-mounted kinematic linkage that allows for large, three-dimensional movements to incorporate the shoulder and elbow. The ReJoyce also includes a screen that allows for the inclusion of gaming interfaces. The FitMi is composed of two sensorized pucks that can be placed on the table or floor, worn on the body, or held in one’s hand to guide and monitor arm, leg, and hand-based exercises. GripAble is handgrip device with a compliant shell. The device is instrumented with a load cell and inertial measurement unit for measuring grip force and wrist/hand movements, respectively, and uses these as inputs to a variety of gaming interfaces where grip force or wrist movement controls the position of a digital object (Broderick et al., 2021; Broderick et al., 2023; Lotay et al., 2019; Mace et al., 2017).
Owing to the higher rate of commercialization of these custom devices for augmenting feedback, many of these devices have also been examined in randomized controlled trials that have shown promising improvements in impairment and function. A randomized controlled study with the Smart Glove found that it improved UE-FM and Jebsen-Taylor hand function test (hand disability) following the intervention and more than a control group that received conventional OT, as well as increased oxygenated hemoglobin concentration in the affected primary sensorimotor cortex (Shin et al., 2022). A randomized controlled study with the MusicGlove found that it improved hand disability as measured by the Box and Block and Nine Hole Peg test more than conventional training (Friedman et al., 2014). A randomized controlled study with FitMi found that UE-FM scores increased more than conventional, home-based therapy (Swanson et al., 2023). A randomized controlled study with the ReJoyce found that participants improved by clinical measures of impairment, disability, and cognition, but no more than dose-matched conventional OT.
Devices for Directing and Enhancing Movement Quality
In addition, there are devices that provide augmented feedback by encouraging the patient to practice movement qualities that are desirable in the context of rehabilitation. Examples of these devices include a pair of hip extension enhancing eyeglasses (Hinton et al., 2024; Hinton et al., 2023), vibrotactile shoe inserts (Afzal et al., 2019), and the Tele-REINVENT (Marin-Pardo et al., 2024; Marin-Pardo et al., 2021).
The hip extension eyeglasses are worn by a stroke survivor during overground walking that display real-time feedback of their hip extension angle as well as a target extension angle that the participant should aim to reach with each stride. The motivation for this device is inspired by previous studies in stroke survivors that have shown that increasing the trailing limb angle (i.e., the angle of the vector connecting the hip and ankle joint centers) leads to a greater propulsive force on the more impaired limb (Genthe et al., 2018; Lewek et al., 2018; Lewek and Sawicki, 2019), which has been shown to be an excellent indicator of walking function in stroke survivors (Roelker et al., 2019). An unrandomized cross-over study with this device found that the device led to sustained increases in hip extension and paretic propulsive force as compared to an equal bout of normal walking, although it is important to note that participants in this study were explicitly told to use the walking strategy from training in their post-test. Therefore, it is unclear if this effect is due to the device or deliberate gait adaptation following instruction.
The vibrotactile shoe inserts are instrumented shoe inserts with different programmable vibration paradigms, including vibration magnitude proportional to stride asymmetry so that users aim to reduce their asymmetry during overground walking. Previous studies with these inserts have found that they lead to online improvements in step length symmetry (Afzal et al., 2019). The same developers of these inserts also developed the previously mentioned Haptic Cane Device, which is capable of providing assistance as well as kinesthetic (force-based) feedback to the user to encourage them to use their more-impaired lower extremity.
The Tele-REINVENT is an electromyography (EMG) biofeedback system that uses low-cost, off-the-shelf sensors to integrate EMG into a variety of gamified activities (Marin-Pardo et al., 2021). An uncontrolled in-home study with this system found that it increased wrist ROM but had no effect on clinical measures. It is important to note that many of the activity-based platforms mentioned in the previous paragraph also incorporate this form of feedback. For instance, the Smart Glove, FitMi, and NeuroBall include activities that provide kinematic feedback and encourage users to challenge their own excursion (e.g., increasing extension of the wrist).
Discussion
Low-cost devices are an emerging approach to increase the effectiveness of stroke rehabilitation as well as increase accessibility to cutting-edge rehabilitation technologies. Hence, we examined the different strategies that researchers use to lower the cost of their devices. In this review, we highlighted 37 devices that were designed with the specific motivation of being low-cost relative to existing rehabilitation devices and had been evaluated in cohort of stroke survivors (Figure 2). The devices/studies originated from USA, UK, Canada, South Korea, India, Serbia, Turkey, Japan, the Netherlands, Germany, Malaysia, Singapore, China, Brazil, New Zealand, and Israel. The prices of these devices (when available) generally ranged from a few hundred USD (FitMi, MusicGlove, GripAble, TheraBracelet, Neuroball) to several thousand USD (Neofect Smart Board, AbleX systems, eddy current braking leg brace, Reha-Slide), with a few as high as $15,000-40,000 USD (Neofect Smart Glove, SaeboReJoyce, ArmeoBoom, ArmeoSpring). Software solutions using after-market off-the-shelf components (e.g., PlayStation Eyetoy) could be reasonably made for less than $100 USD. The majority of these devices only target the upper extremity or joints of the upper extremity, while fewer target the lower extremity. We synthesized the details of these devices and the studies examining them and extracted three methods by which these devices interface with patients: assistive (devices that assist the user’s motion during training), resistive (devices that resist the user’s motion during training), and augmented feedback (devices that alter sensory feedback during training). Furthermore, we highlighted specific types of rehabilitation technology—active, passive, and semi-passive devices—that assist or resist patient motion with different actuator types. This review provided a state of the art of low-cost technologies for stroke rehabilitation, which we believe will be useful to other researchers aiming to increase accessibility of rehabilitation resources.
Existing low-cost device for stroke rehabilitation binned by their mode of training. It is important to note that some devices appear in multiple columns because they employ multiple modes. Here, Motorized and Stimulation-based devices can be categorized as “Active” devices because they add energy to the user-device system. Gravity Compensation, Inertia-based, Brake, Self-power, Exosuits, and Tone compensation device can be categorized as “Passive” devices because they dissipate or recycle the user’s energy but cannot be controlled in real-time. Gamified environments and Biofeedback devices can be categorized as “Augmented Feedback” devices because they do not add or dissipate energy; they only provide visual or haptic cues to increase engagement or encourage correct movements.
We found several useful pieces of information regarding the effectiveness of low-cost devices in stroke rehabilitation. First, it appears that low-cost devices for stroke rehabilitation are at least as effective as standard clinical practices at improving extremity function. This is evidenced by the plethora of studies discussed above that employed randomized controlled designs and found that the device was as effective (Budhota et al., 2021; Hesse et al., 2008; House et al., 2017; Housman et al., 2009; Kim et al., 2018; Ozen et al., 2021; Park et al., 2019) or more effective (da Silva Ribeiro et al., 2015; Friedman et al., 2014; Kim et al., 2018; Ozawa et al., 2013; Scronce et al., 2023; Shin et al., 2022; Swanson et al., 2023; Tomic et al., 2017; Yavuzer et al., 2008) in reducing extremity impairment and/or disability and increasing community participation. This is an encouraging finding for researchers hoping to continue developing these devices. Furthermore, it is important to note that many of these studies examined their device with dose-matched control interventions. As mentioned previously, a critical advantage of low-cost devices is their potential to improve dosage through increased clinician efficiency, home-based therapy, and patient engagement. Therefore, no differences between low-cost devices and standard-of-care could be interpreted as an advantage for low-cost devices when accounting for their potential to increase dosage.
This review also highlighted common themes in the design of these low-cost devices. Specifically, augmented feedback devices that provide activity/game-based environments are one of the most common types of low-cost device for stroke rehabilitation, as over one-third of the devices shown in Table 1 exclusively provide augmented feedback. Furthermore, this type of device has the greatest penetration into the market, exemplified by MusicGlove, SaeboReJoyce, FitMi, and others. This popularity among researchers and industry members is understandable because assistance/resistance requires an interface for device interaction, sensors to detect movement, an actuator to generate the force, and a structure to transmit the force to the user with minimal losses, all of which can increase cost. An augmented feedback device, on the other hand, only requires the interface and the sensors, and therefore has a distinct advantage over these other device types in terms of decreasing costs. Indeed, many active, passive, and semi-passive devices integrate augmented feedback into training because they already use a computer to sense and react to the user’s movement, therefore adding augmented feedback requires no additional hardware (Afzal et al., 2018; Budhota et al., 2021; House et al., 2016b; Kilbride et al., 2022; Ozawa et al., 2013; Park et al., 2019; Tomic et al., 2017).
Interestingly, very few custom devices (Afzal et al., 2018; Hinton et al., 2024; Marin-Pardo et al., 2024) used augmented feedback to alter the quality of the patient’s movements during task-oriented practice. This could be due to an interaction between most devices being designed for upper extremity rehabilitation and the complexity in designing feedback that is intuitive, improves movement quality, and can be applied to general usage of the upper extremity. Indeed, the hip extension eyeglasses, vibrotactile inserts, and haptic cane device all provide feedback only to improve gait. As such, future augmented feedback devices could consider methods of improving upper extremity movement quality (e.g., less trunk compensation, better joint individuation, etc.) as a means of improving device effectiveness. It is important to note that low-cost, off-the-shelf augmented feedback devices mentioned in this review have been used elsewhere to improve upper extremity movement quality in stroke survivors (e.g., Microsoft Kinect) (Ballester et al., 2015; Cameirao et al., 2012; Cameirao et al., 2010). However, these studies were not included in the current review due to their lack of focus on these systems as low-cost rehabilitation devices.
Another common theme in the design of low-cost devices was that many researchers opted to design passive systems instead of active ones. Furthermore, many of these passive devices focused on gravitational forces as the channel of assistance/resistance. For instance, the assistive passive devices, which constituted the second most common form of low-cost device, typically focused on reducing gravitational loads via passive limb support to facilitate the integration of the limb into task-oriented practice. Similarly, the most common method of resistance was using tilted surfaces or masses to increase the gravitational forces that must be overwhelmed to accomplish functional activities. Indeed, of the 37 devices considered in this review, 11 focused on decreasing or increasing gravitational loads as opposed to 5 that used motors to assist/resist the patient. As discussed in the previous sections, this finding is understandable, as electrical motors that can exert enough torque to assist or resist movement require additional equipment and safety considerations that can increase costs. Indeed, of the five motorized devices included in this review, only one device included more than one motor.
Surprisingly, despite the research community's recognition that motors increase costs, existing research in passive and semi-passive devices gives an incomplete picture of the potential of these devices in stroke rehabilitation. To recall, passive devices use uncontrolled passive force generation elements (e.g., elastic bands/springs) to produce patient-device interaction forces, and semi-passive devices use controllable passive force generation elements (e.g., computer-controlled brakes) to control forces in real-time. Of the sixteen passive devices examined in this review, the majority target the upper extremity, while only three targeted the lower extremity (Ventura et al., 2019; Washabaugh and Krishnan, 2018; Yao et al., 2021), and only two targeted the hand and finger function (Casas et al., 2021; Jang et al., 2016). There was only one semi-passive device included in this review (Ozawa et al., 2013), and although it is a clever design, it is only capable of providing simple resistance to a user’s motion, which is far less functionality than similar motorized devices (Marchal-Crespo and Reinkensmeyer, 2009).
It is important to note that the selection criteria used in this review to extract low-cost devices from literature may have influenced our analysis in a few ways. First, the scope of the review was limited to devices that have been designed with the explicit intent of reducing cost of rehabilitation devices and has been evaluated in at least 6 strokes survivors. This necessarily leaves out large number of devices that have been developed to reduce cost but evaluate their device in uninjured controls, less than 6 stroke survivors, and/or other patient populations. These restrictions were necessary to feasibly produce this review, but the result is that some valid engineering approaches to reducing device cost may not have been covered. Surprisingly, this criterion also eliminated a number of devices that are marketed as low-cost stroke rehabilitation devices (e.g., XR Health, Neuro Rehab VR) but whose use is supported by general evidence rather than studies examining the specific device. This is unfortunate, as it draws into question the validity of these platforms to target stroke-specific deficits. Next, the distinction of low-cost was determined by the authors/inventors of devices included in this review rather than a fixed value threshold. We elected this approach to avoid the risk of biasing results based on a specific definition of “low-cost”, which could be affect by both institutional and national resources, and to avoid issues if cost was not available for laboratory-based prototypes. However, this approach relies on each authors’ definition of “low-cost”, which may be not always be consistent across devices.
In conclusion, we performed a review of previously developed low-cost devices for stroke rehabilitation. In this review, we examined 37 low-cost devices for stroke rehabilitation that demonstrate promising effectiveness for targeting post-stroke extremity deficits. Additionally, the review uncovered that augmented feedback and passive systems are promising avenues for reducing device cost. However, the review also showed that there are several critical gaps in this field. First, most low-cost devices target the upper extremity and very few target the lower extremity or fingers, which are both critical for stroke recovery. Second, there are very few semi-passive devices that have been examined in stroke rehabilitation, and the devices that have been examined provide functionality that is considerably reduced compared to their motorized counterparts. As such, future research in low-cost stroke rehabilitation devices should aim to expand knowledge in these areas.
Supplemental Material
sj-docx-1-rnn-10.1177_09226028261423567 - Supplemental material for Low-cost Devices for Stroke Rehabilitation: A Review of Approaches, Designs, and Evidence
Supplemental material, sj-docx-1-rnn-10.1177_09226028261423567 for Low-cost Devices for Stroke Rehabilitation: A Review of Approaches, Designs, and Evidence by Thomas E Augenstein PhD and Chandramouli Krishnan PT PhD in Restorative Neurology and Neuroscience
Footnotes
Ethical Considerations
No human subjects were recruited for the purposes of this study; therefore, this section is not applicable.
Consent to Participate
No human subjects were recruited for the purposes of this study; therefore, this section is not applicable.
Consent for Publication
Not Applicable
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Science Foundation (DGE #1256260, 1804053), the Eunice Kennedy Shiver National Institute of Child Health and Human Development / National Institutes of Health (R41-HD111289), and the University of Michigan Rackham Predoctoral Fellowhsip.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability
The data supporting the findings of this study are available on reasonable request from the corresponding author.
Supplemental Material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
