Welcome to SCIRE Professional

Recently, robotic devices were developed as a non-invasive solution to enhance intact motor pathways or manipulate the upper limbs for functional improvement (Capello et al., 2018). A number of different robotics are currently used for rehabilitation and they can be classified based on the type of robot, actuation method (energy source, e.g. electric motor), form of transmission (transfer of motion, e.g. cables) and sensors used (Yue et al., 2017). The two most common types of robotic devices used include end-effectors and exoskeletons (Yue et al., 2017). End-effectors are attached to the end of a robotic arm (e.g. robotic hand) and are designed to interact with the environment, externally to the patient (Yue et al., 2017). In contrast, exoskeletons are worn by the patient and include mechanical joints that align to the subject’s own joints, which assist the impaired user to move their own upper limbs (Sicuri et al., 2014; Yue et al., 2017; Capello et al., 2018). Importantly, both types of robotic devices may be used to deliver high quality and high volume repetitions. It was recently suggested that repetitive movement exercise may promote functional recovery through the enhancement of adaptive plasticity (Frullo et al., 2017; Capello et al., 2018). A large body of literature has described the efficacy of robot-assisted rehabilitation for recovery of upper extremity motor function in stroke patients (Lo et al., 2010; Klamroth-Marganska et al., 2014; Frullo et al., 2017). However, there is a paucity of data on the efficacy of robot-assisted rehabilitation for recovery of upper extremity motor function in SCI.

The methodological details and results from nine studies are listed in Table 7.

Table 7 Upper Limb Robotics Interventions

Author Year


Research Design


Total Sample Size

Frullo et al. 2017




Population: Assist-as-needed (AAN) robotic controller: Mean age=53.5 yr; Gender: not reported; Time since injury: 16 yr; Level of injury: C3 – C6.

Subject-triggered (ST) robotic controller: Mean age=53.5 yr; Gender: not reported; Time since injury: 16 yr; Level of injury: C3 – C6.

Intervention: Participants were assigned to AAN or ST robotic controller groups. One wk after the last baseline visit, subjects started a program of robotic training, in ten 90-min long sessions, spread over a period of three to four wk. Outcome measures were assessed at baseline, one wk, two wk, and two mo after treatment.

Outcome Measures: Action Research Arm Test (ARAT); Modified Ashworth Scale (MAS); Grip Pinch Strength assessment (GPS); GRASSP.

1.     No significant difference was observed in the ARAT, MAS, GPS, or GRASSP scores or between groups (p>0.05).

2.     The AAN robotic controller demonstrated gradual improvement in movement quality over the ST robotic controller.

Capello et al. 2018




Population: Mean age=49.8 yr; Gender: males=8, females=1; Time since injury: 26.9 yr; Level of injury: C4-C7, tetraplegia=9; Severity of injury: not reported.

Intervention: Tetraplegic patients were administered a hand function test to assess the functionality of a soft robotic glove. Outcome measures were assessed at baseline without the assistive glove and once while wearing the assistive glove.

Outcome Measures: Hand function during ADL tasks (Toronto Rehabilitation Institute Hand Function Test (TRI-HFT)); Object manipulation; Lift force.

1.     The soft robotic glove significantly improved key hand functions to manipulate ADL objects and the mean score between baseline and assisted condition across all TRI-HFT categories (p<0.05).

2.     Lift force increased significantly when using the assistive soft robotic glove (p<0.05).

Kim et al. 2017




Population: Mean age=33 yr; Gender: males=4; Time since injury: 12 yr; Level of injury: C5 – C6; Severity of injury: AISA A=2, B=2.

Intervention: Participants compared writing performance using a new hand assist device (GRIPIT) to writing performance with a conventional penholder and their own hand without any device. Outcome measures were assessed at baseline and while using each assistive device.

Outcome Measures: Quantitative outcomes: Accuracy of writing; Solidity of writing; Qualitative outcomes: Appearance; Portability; Difficulty of wearing; Difficulty of grasping; Writing sensation; Fatigability; Legibility.

1.     Quantitative results showed that GRIPT users perform significantly better on accuracy and solidity of writing than conventional pen holders or with their own hand (p<0.05).

2.     Qualitative results showed that GRIPIT has advantages for writing sensation, fatigability, and legibility; Participants found it more difficult to wear than a conventional pen holder; No difference was observed in portability and difficulty grasping (p>0.05).


Backus et al. 2014




Population: Mean age: 40.5±13.0 yr; Gender: males=8, females=2; Level of injury: C2-C3=3, C4-C7=7; Mean ASIA motor score: 15.8±3.9; Mean time since injury: 3.0±1.1 yr.

Intervention: Test effect of assisted movement with enhanced sensation (AMES) using vibration to antagonist muscle to reduce impairments and restore upper limb function in people with incomplete tetraplegia. Two or three sessions over 9-13 wk per participant.

Outcome Measures: Strength and active motion tests on the AMES device, International Standards for the Neurological Classification of SCI (ISNCSCI) motor and sensory examinations, Modified Ashworth Scale (MAS), grasp and release test (GRT), Van Lieshout Test (VLT), Capabilities of Upper Extremity questionnaire (CUE).

1.     No significant change in MAS scores (p=0.371) or ISNCSCI scores (p=0.299 for motor, p=0.459 for sensory-light tough, p=0.343 for sensory-pin prick).

2.     Strength test scores increased significantly for MCP extension (p≤0.01) and flexion (p≤0.05) and for wrist extension (p≤0.001) and flexion (p≤0.01).

3.     Active motion test scores increased significantly for MCP joints (p≤0.001) and wrist (p≤0.001).

4.     Out of GRT, VLT and CUE scores, only GRT scores were significantly improved after training and slightly between post treatment and 3-mo post treatment (p=0.025).


Cortes et al. 2013




Population: Mean age: 44.8±16.3 yr; Gender: males=8, females=2; Level of injury: C4-C6=10; Severity of injury: AIS-A complete=3, AIS-B incomplete=4, AIS-C incomplete=1, AIS-D incomplete=2; Mean time since injury: 4.7±2.5 yr.

Intervention: Chronic tetraplegic SCI patients participated in a 6-wk wrist-robot training protocol (1hr/day, 3 times/wk) to evaluate feasibility, safety and effectiveness on upper limb.

Outcome Measures: Motor performance, Corticospinal excitability, Upper extremity Motor score (UEMS), Visual Analogue Scale (VAS), Modified Ashworth Scale (MAS), resting motor threshold (RMT), Motor evoked potential (MEP) amplitude and latency at rest, MEP facilitation.

1.     Significant improvements in aim and smoothness (p=0.03).

2.     No changes in deviation, mean speed, peak speed and duration of movement was found.

3.     No changes in motor strength of trained right arm (p=0.4) or untrained left arm (p=0.41).

4.     No significant changes in MAS of either arm (p>0.05 for both).

5.     No significant changes in pain levels after training (p=0.99).

6.     There were no changes in any neurophysiological parameters after the 6-wks of training (p>0.05).

7.     Strong positive correlation between change in smoothness according to the initial spasticity level (R2=0.403); change in aim was positively correlated with initial spasticity in trained arm (R2=0.123)

8.     Initial UEMS and MEP amplitude had no correlation with the change on smoothness and aim.

Tigra et al. 2018




Population: Mean age=36.4 yr; Gender: not reported; Time since injury: 10.7 yr; Level of injury: C5 – C7; Severity of injury: AISA A=4, B=0, C=1, D=0.

Intervention: Participants piloted a newly developed assistive device (human-machine interface) for hand grip function that utilizes EMG signals from selected muscles to operate a robot hand.

Outcome Measures: Voluntary muscle contraction (EMG); Hand grasping.

1.     Although no statistics were reported, all subjects were able to individually contract the tested muscles on demand for at least 7 s (indicated by EMG), except for one participant with no voluntary contraction. EMG signals were turned into functional commands to pilot the hand.

2.     The tasks (holding an object in the robot hand for 5 s, open hand, palmar pinch and key grip) were successfully achieved with each tested muscle, however, no statistics were reported.

Popovic et al., 1999




Population: Mean age: 26.5 yr; Level of injury: C5-C7; Severity of injury: complete=10, incomplete=12; Length of experience with device: ≥6 mo.

Intervention: Subjects utilized a bionic glove to complete functional testing of quantitative and qualitative outcome measures.

Outcome Measures: Quadriplegia Index of Function (QIF), Functional Independence Measure (FIM), Upper Extremity Function Test, Goniometric Measurements.

1.     QIF: mean was 19.0±6.5 at the beginning; at the end 28.4±5.2, improvement of 49.5%.

2.     FIM: 63.8±10.4 at the beginning; 79.0±8.9 after six mo. When three clients excluded who had 120 points on FIM scores were beginning 44.4±13.5 and 64.8±16.6 after six mo (increase of 20.4 points/46%).

3.     Functional task completion: six subjects continued to use the device. On average, 75% of the functions were performed better after six mo of use. 6/12 (50%) did not continue to use the device. C6-C7 individuals may find the device beneficial enough to use it as an assistive device.

4.     Technical improvements, specifically cosmetics, positioning of the electrodes, donning/doffing, should increase the number of regular users.

5.     Best candidates are individuals with complete C6-C7 tetraplegia.

6.     FIM score between 25-50 (up to 75), QIF between 0-13 (up to 27), are motivated to use it, can demonstrate efficient grasp.

Prochazka et al., 1997




Population: Age: 22-42 yr; Gender: males=8, females=1; Level of injury: C6-C7; Time since injury: 16 mo–22 yr.

Intervention: Use of bionic glove.

Outcome Measures: Mean peak force of tenodesis grasp, Qualitative ratings of manual tasks.

1.     Mean peak force of tenodesis grasp in the nine subjects increased from 2.6 N±3.8 N (passive) to 11.3 N±7.4 N (glove active), significant than peak passive force (p=0.0064, t-test), and significant at end of fifth grasp 6.8 N±4.2 N, p=0.0064, Mann-Whitney rank sum test.

2.     Most manual tasks improved significantly with the use of the glove.

Coignard et al. 2013




Population: Injury Group (n=29): Mean age=37.8±13.3 yr; Injury etiology: spinal cord=23, post-stroke locked in syndrome=2, arthrogryposis=1, quadruple amputee=1, cerebral palsy=1, spinal muscular atrophy=1; Controls (n=34): Mean age=32.4±11.2 yr.

Intervention: No intervention. To evaluate the reliability and functional acceptability of the ‘‘Synthetic Autonomous Majordomo’’ (SAM) robotic aid system in a domestic environment using three multi-step scenarios: selection of the room in which the object to be retrieved was located, selection of the object to be retrieved, the grasping of the object itself and the robot’s return to the user with the object.

Outcome Measures: Selection time (time between task’s “start” command and room/object selection click), Number of failures, Qualitative questionnaire.

1.      No significant difference between scenarios 1 and 2 in room/object selection, validation times and number of failures for controls and patients (p>0.05).

2.      Statistically significant difference between scenario 2 and 3 in object selection time for controls and patients (p<0.05) but not for number of object selection failures (p>0.05).

3.      Patients took significantly longer to select the room and the object than the controls did (for room selection in scenarios 1 and 3 and for object selection in all three scenarios) (p<0.05).

4.      No significant patient versus control differences in the number of failures (p>0.05).

5.      Experience of computer use had significantly affected speed of task for patients in scenario 3 (p<0.05) and controls in all scenarios (p<0.05).

6.      Overall, the robot was found to be acceptable by both patients and control participants.


The field of robotic devices for SCI rehabilitation is constantly evolving as technology advances. As a result of this, the majority of articles published in this area focus on testing newly designed robotic devices via non-randomized pilot studies that contain small sample sizes. Accordingly, it is difficult to draw any definitive conclusions about the efficacy of robotic rehabilitation itself. It is more appropriate to discuss emerging trends with specific types of robotic devices for SCI rehabilitation.

Several studies examined the feasibility and efficacy of robotic exoskeletons. All of the studies found that use of a robotic exoskeleton is feasible, however, the real world functionality of it may be limited and hard to use based on individual functioning. For example, one study found that use of a bionic glove was only successful in patients that had voluntary control over their wrist, while another found that at home use of the device may be impractical. In contrast, other studies conducted using different types of exoskeletons (e.g. GRIPIT and a soft robotic based glove) found significant improvements in writing and hand function while wearing the device. GThe efficacy of exoskeleton use is controversial and may vary depending on the type of exoskeleton used and the overall functioning of the patient.

Only a few studies examined the feasibility and efficacy of an end-effector robotic device. However, all of the studies demonstrated improvements in upper extremity function while using the device. It should be noted that end effectors are robotic devices aimed at replacing upper extremity function instead of rehabilitating the patient. With the current technology available, robotic end-effectors are often cumbersome and large with complex interfaces. As such, Coignard and colleagues (2013) found that use of one at home is much less feasible than in a clinical setting. At present, this makes the feasibility of robotic end-effector rehabilitation fairly low. As technology advances, robotic end-effectors may evolve to be more adaptable in an at-home setting. Future research should focus on the long-term efficiacy, as well as determining usability through functional impact questionnaires (e.g. FIM and ADL).


There is level 2 evidence (from one prospective controlled study; Frullo et al. 2017) that subject-adaptive upper extremity robotic exoskeleton therapy is feasible, however, no gains in arm function were observed.

There is level 4 evidence (from one pre-post study; Capello et al. 2018) that use of a fabric-based soft robotic glove significantly improves hand function when completing activities of daily living in individuals with SCI.

There is level 4 evidence (from one pre-post study; Kim et al. 2017) that the GRIPIT exoskeleton quantitatively and qualitatively improves writing when compared to conventional pen holders, although it is more difficult to wear.

There is level 4 evidence (from two pre-post studies; Backus et al., 2014; Cortes et al., 2013) that an end effector can be safely used in patients with tetraplegia to significantly improve upper limb function.

There is level 4 evidence (from one post-test study: Tigra et al., 2018) that an end effector robotic device may improve hand grasping function in individuals with SCI.

There is level 4 evidence (from two case series; Popovic et al., 1999; Prochazka et al., 1997) that the Bionic Glove increases motor and upper limb function in individuals with SCI.

There is level 5 evidence (from one observational study; Coignard et al., 2013) that in a home environment the functionality of an end effector may be limited.

Upper extremity robotics improve hand function in individuals who have suffered upper limb paralysis following a spinal cord injury. However, further research is necessary to determine the efficacy of upper extremity robotic exoskeletons as part of a robotic rehabilitation program.