916 IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 27, NO.

5, MAY 2019

Hand Extension Robot Orthosis (HERO) Glove:

Development and Testing With Stroke Survivors
With Severe Hand Impairment
Aaron Yurkewich , Student Member, IEEE , Debbie Hebert, Rosalie H. Wang,
and Alex Mihailidis

Abstract — The hand extension robot orthosis (HERO) I. I NTRODUCTION

glove was iteratively designed with occupational therapists
and stroke survivors to enable stroke survivors with severe
hand impairment to grasp and stabilize everyday objects,
while being portable, lightweight, and easy to set up and
F IFTEEN million individuals worldwide experience a
stroke each year with 50,000 of these cases occurring
in Canada [1]. Approximately two-thirds of these individ-
use. The robot consists of a batting glove with artificial uals will experience neurological deficit [2] and half will
tendons embedded into the glove’s fingers. The tendons
are pulled and pushed by a linear actuator to extend and never fully recover the hand function required to perform
flex the fingers. The robot’s finger extension and grasp activities of daily living independently [3]. High-intensity,
assistance are automated using inertial measurement unit personalized therapy is needed for stroke survivors to regain
signal thresholds. Five stroke survivors (Chedoke McMaster their hand range of motion (ROM), strength and coordina-
Stroke Assessment – Stage of Hand 1-3) put on the HERO tion, and translate recovery gains into daily task indepen-
Glove in 1-3 minutes, with assistance. The stroke survivors
performed significantly better on the Box and Block Test dence [4], [5]. The coaching and motion assistance required
(2.8 more blocks transferred, p < 0.01) while wearing is time and resource intensive, making it difficult for therapy
the HERO Glove than when not wearing the glove. Four clinics to supply at the appropriate intensity [6], [7]. After
stroke survivors could only transfer blocks while wearing their therapy programs are complete many stroke survivors
the HERO Glove. The HERO Glove enabled these stroke still struggle to extend their fingers to prepare for a grasp
survivors to more fully extend their index finger (an increase
of 97.5◦ , p < 0.01) and three of five stroke survivors were and grip with enough strength to stabilize objects, thereby
better able grasp a water bottle. Therapists and stroke making the hand difficult to integrate into daily activities at
survivors suggested increasing the HERO Glove’s grip home.
force assistance and valued the glove’s portability, light- Therapists and stroke survivors have steered technology
weight design and potential usefulness in assisting with designers to create wearable robotic hand orthoses that can
task-based therapy.
make hand therapy more engaging and reduce the activation
Index Terms — Rehabilitation robotics, assistive technol- barrier to performing unsupervised exercises at home [8]–[10].
ogy, occupational medicine, hand therapy, stroke, wear- Current devices have been developed for in clinic use
ables.
and have shown recovery benefits close to minimum clin-
Manuscript received October 27, 2018; revised February 20, 2019; ically meaningful differences (i.e. 1 to 6 points on the
accepted April 5, 2019. Date of publication April 11, 2019; date of Fugl-Meyer Assessment-Upper Extremity (FMA-UE) and
current version May 9, 2019. This work was supported in part by the
University of Toronto, in part by the Toronto Rehabilitation Institute (TRI), Motor Activity Log) [11]–[16]. By improving the porta-
in part by the Natural Sciences and Engineering Research Council bility, affordability and ease of use of wearable robotic
of Canada (NSERC), in part by the Canadian Partnership for Stroke hand orthoses, they could be more easily integrated into
Recovery (CPSR), and in part by the AGE-WELL NCE Inc., through
the Networks of Centers of Excellence Program. (Corresponding author: home therapy programs to monitor active movement, assist
Aaron Yurkewich.) directly in daily activities and increase neuromuscular recovery
A. Yurkewich is with the Institute of Biomaterials and Biomedical [17]–[19].
Engineering, University of Toronto, Toronto, ON M5S 3G9, Canada, and
also with the Toronto Rehabilitation Institute (TRI), Toronto, ON M5G 2A2, A small number of wearable hand robots intended for
Canada (e-mail: [email protected]). home use have been evaluated with people with affected
D. Hebert is with the Rocket Family Upper Extremity Clinic, TRI (e-mail: hands. These robots’ assistive capacity has been shown to
[email protected]).
R. Wang is with the Department of Occupational Science and Occu- enhance hand function and performance on activities of daily
pational Therapy, University of Toronto, Toronto, ON M5S 3G9, Canada living (ADLs). A summary of these and other recently-
(e-mail: [email protected]). developed wearable hand robots, including their components
A. Mihailidis is with the Department of Occupational Science and
Occupational Therapy, Institute of Biomaterials and Biomedical Engi- and trial efficacy, is presented in Supplementary Table I. For
neering, University of Toronto, Toronto, ON M5S 3G9, Canada (e-mail: stroke survivors with moderately-affected upper extremities,
[email protected]). Peters et al. [20] showed that elbow, wrist, and finger assis-
This paper has supplementary downloadable material available at
http://ieeexplore.ieee.org, provided by the authors. tance improved FMA-UE scores by 13% as well as cup and
Digital Object Identifier 10.1109/TNSRE.2019.2910011 utensil grasping. For stroke survivors with severely-affected

This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/
YURKEWICH et al.: HERO GLOVE: DEVELOPMENT AND TESTING WITH STROKE SURVIVORS 917

Fig. 1. The HERO Glove consists of (a) cable tie pawls, (b) a linear actuator, (c) a 9V battery pack, 9V battery and microcontroller with an inertial
measurement unit, (d) buttons to control manual mode and select between the manual and automated modes, (e) a batting glove, (f) cable tie
tendons, (g) finger thimbles, and (h) an open palm. The HERO Glove is shown extending (left) and flexing (top, right) the relaxed fingers and
thumb.

hands, Park et al. [21] showed that a 40N extension force could in enhancing daily task independence and provide design
extend low-tone but not high-tone fingers, making essential guidance for wearable robotic hand orthosis designers.
skills like cylindrical grasping difficult [8]. In addition, stroke
survivors can have weakened grip strength and may need at II. M ETHODS
least 15N of palmar or pinch grasp assistance to complete daily
tasks, as demonstrated by Cappello et al. [22] for spinal cord A. HERO Glove Design
injury survivors. As a result, multiple sizeable actuators and The HERO Glove, shown in Fig. 1, was iteratively designed
energy storage units have been integrated into these robots, and tested with occupational therapists specialized in stroke
which are not aesthetically pleasing and increase the weight therapy, engineering students and two chronic stroke survivors
such that an arm support is required [9] or require additional with severe hand impairment [Stage 3 and Stage 1 hands
cabling and donning processes for back, belt or wheelchair (out of 7) - Chedoke McMaster Stroke Assessment [27]; level
mounting that may reduce usability [22]–[26]. Buttons, elec- 2 tone (out of 4) - Modified Modified Ashworth Scale [28]],
tromyography (EMG), electroencephalography (EEG), voice both of whom presented with high finger and wrist tone and
and vision have been used to sense the user’s intent, showed no active finger extension. Initial requirements for the
in order to trigger assistance and motivate spontaneous wearable hand robot were generated through bi-weekly meet-
use of the affected upper extremity. However, the accuracy ings between therapists and engineers and conversations and
in detecting the user’s intent during robot-assisted trials robot testing with the chronic stroke survivors. Previous inter-
has only been reported in Soekadar et al. (with a 16.3% view findings were used to prompt these conversations [8], [9].
false-positive rate) [25] and stroke survivors with severely Quantitative specifications, shown in Table I, were assigned for
impaired upper extremities are often excluded from studies each requirement after discussing the specifications of previous
because the sensing modality cannot accurately detect their wearable hand robots.
intent [20]. The HERO Glove transmits extension and flexion forces to
This article details the design and evaluation process taken the index and middle finger and thumb through cable ties. The
by our transdisciplinary team of researchers, therapists and ends of the cable ties are fixed to the fingertips of a batting
stroke survivors at the Toronto Rehabilitation Institute - Uni- glove (Mizuno Supreme, Men’s Large) and slide through cable
versity Health Network (TRI-UHN) to develop the Hand guides fixed to the dorsal side of the glove. The cable ties
Extension Robot Orthosis (HERO) Glove. This glove has been are actuated by push-pull forces from a single linear screw-
designed to reduce barriers to using the stroke hand in daily drive servo actuator (Actuonix, L12-R, 210:1, 80N max force,
life by enabling stroke survivors with severe hand impairment 50mm stroke length) that is mounted on the dorsal surface
to grasp and stabilize everyday objects through mechanical of the glove in-line with the two proximal sets of cable
assistance of finger and thumb extension and flexion. Key guides. When the actuator extends, the cable ties pull on the
attributes of the HERO Glove are its portability, light weight, fingertips of the batting glove and apply a straightening force
ease of donning, use of affordable components and inertial on the dorsal side of the fingers. When the actuator contracts,
measurement unit (IMU) triggered control method for one- the cable ties apply a bending force to the fingers because the
handed use. The HERO Glove’s motion assistance capabilities batting glove restrains their axial motion. The actuator is non-
are validated with stroke survivors to understand its efficacy backdrivable, which conserves power when extension or grip
918 IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 27, NO. 5, MAY 2019

assistance is required for long periods of time. The actuator TABLE I

and the onboard microcontroller (tinyTILE Intel Curie) are T HERAPIST AND S TROKE S URVIVOR D ESIGN S PECIFICATIONS
AND THE HERO G LOVE ’ S C APABILITIES
powered by a 9 Volt battery (Energizer, Rechargeable NiMH).
Separate HERO Gloves were created for the left and right
hands to make the study more inclusive. The glove’s fingertips
have high-friction silicone webbing for added grip. Perforated
rubber thimbles (Staples Fingertips) were glued inside the
finger tips to increase comfort and make finger insertion easier
while minimizing any additional loss of fingertip sensation.
The HERO Glove’s specifications are shown in Table I.

B. HERO Glove Control Strategy

The user operates the glove using the two control
modes shown in Supplementary Figure 1 and available at
github.com/drossos/HERO-robot. The user can press the phys-
ical button to extend and flex the fingers or have the micro-
controller automatically trigger assistance when the gyroscope
reading from the inertial measurement unit (IMU) reaches a
threshold. The purpose of the IMU-triggered automated mode
is to keep the unaffected hand free and to motivate arm and
wrist use as the IMU is positioned distal to the wrist. To deter-
mine the threshold values for triggering motion assistance,
four able-bodied participants (age 18-35; 3 female) simulated
moving their arm slowly while wearing the HERO Glove to
perform the Box and Block Test [29]. Their gyroscope data
provided a higher signal to noise ratio than their accelerometer
data and a threshold value of 0.23◦/s was selected. The glove
alternates between triggering flexion and extension assistance
when the user moves their hand with an angular velocity above
the threshold and then stops moving for 0.8 seconds. After
the actuator is triggered to move, there is a two second delay
before the actuator can trigger again so the user can adjust their
hand orientation during the grasp. This control scheme allowed
users to reach for a block, stop for the block to be grasped, lift
the block over the barrier and then stop to drop the block. The
four able-bodied participants transferred 7, 4, 5 and 7 blocks
in one minute using the right-handed HERO Glove with their
hand relaxed. The actuator was correctly triggered 46 times
(23 flexion; 23 extension), never triggered when undesired (0%
false positives) and triggered later than desired twice (4% false
negatives). This control scheme also allowed users to reach for
the water bottle, stop to grasp it, lift up and put down the water
bottle and then stop to release the water bottle.

C. Design Decisions Motivated by Iterative Testing

The HERO Glove was developed after creating and testing
the intermediate prototype, shown in Supplementary Figure 2.
The researchers assisted the stroke survivors to don, operate,
and doff the intermediate prototype. Testing the intermediate
prototype identified motion assistance and usability issues that the dorsal actuator; however, this will reduce the extension
motivated three key design iterations. First, cable ties were produced for individuals with high tone. Currently, the ring
added for extension assistance in place of fishing wire. This and little fingers are gloved only for aesthetics. Second,
enhanced comfort and alignment by distributing the force the palm of the glove, the wrist brace and the palmar actuator
applied to the dorsal side of the finger and resisting twisting were removed to make the glove easier to put on, as in
and out-of-plane bending. Cable ties were only positioned on Yap et al. [23] and Polygerinos et al. [30], once it was
the index and middle finger and thumb in order to maximize discovered that the dorsal cable ties could provide some flexion
force. Ring and little finger tendons can be attached to assistance, similar to Nycz et al. [26] and Gandolla et al. [31].
YURKEWICH et al.: HERO GLOVE: DEVELOPMENT AND TESTING WITH STROKE SURVIVORS 919

This modification was necessary because the stroke survivors bend angle of the index finger metacarpophalangeal (MCP),
required 15 minutes of assistance to put on the intermediate PIP, and distal interphalangeal (DIP) joints using a finger
prototype and this was deemed unacceptable for in clinic or at goniometer (JAMAR) in four positions, passive extension,
home use. Removing these components also reduced the size, active flexion, active extension and then passive flexion. Only
weight and cost of the glove and exposed the user’s palm the index finger was measured in order to minimize the length
to avoid blocking sensation. The design tradeoff was that of each study session. The term “passive” refers to when the
the intermediate prototype produced over 2N of pinch force participant was asked to relax their hand for the researcher to
and 17N of grip force for the four able-bodied participants move and “active” refers to when the participant was asked to
with their hands relaxed, while the HERO Glove moved extend or flex their fingers without assistance. Active extension
the relaxed hand into a pinch posture but did not generate was calculated by summing the joint angles at active extension.
force. It was assumed that the stroke survivors’ tone could Passive ROM was calculated by subtracting the passive exten-
stabilize objects so easing the donning process was a higher sion joint angles from the passive flexion joint angles. Active
priority than assisting grip strength. Third, the automated ROM was calculated by subtracting the active extension joint
control mode was developed because the participants had angles from the active flexion joint angles. Further ROM mea-
difficulty pressing the physical button while supporting the surement and calculation details are shown in Supplementary
arm. Figure 3. The finger joints were not extended past straight to
avoid potential injury so the maximum extension was 0◦ for
each joint. The fingers were flexed by the researcher until the
D. Participant Recruitment
fingertip met the palm near the MCP joint. The researcher
Observational case studies with stroke participants with stopped applying force if it was painful to the participant.
limited active finger extension were completed to evaluate the Tone and spasticity in the index finger was assessed dur-
HERO Glove’s efficacy in assisting motion and enhancing task ing the passive extension measurements using the Modified
performance. A convenience sample of stroke survivors was Modified Ashworth Scale (MMAS) [28] and Modified Tardieu
recruited by therapist referral for outpatients and the TRI- Scale (MTS) [33].
UHN central recruitment process for inpatients. This study The robot-assisted ROM was measured using the same
was approved by the UHN Institutional Review Board and instruments, arm posture and finger joints as in the unassisted
each participant provided informed consent to participate in ROM measurements. The glove was donned with assistance
the study. Researchers administered the study methods for to ensure proper alignment and the robot extended the fingers
all stroke survivors, after being trained by an occupational to ensure safe operation. Then the participants were asked to
therapist. Outpatients did not receive therapy prior to the study. keep their hand relaxed or to flex or extend their hand as
Inpatients completed scheduled therapy sessions on the same the robot assisted their motion. The researchers measured the
day as the study. finger joint bend angles in four positions, relaxed-hand robot-
Inclusion Criteria: assisted flexion, flexed-hand robot-assisted flexion, relaxed-
• Stroke survivors more than 1 week post-stroke hand robot-assisted extension, and then extended-hand robot-
• Chedoke-McMaster Stroke Assessment Stage of the Hand assisted extension. The extended-hand robot-assisted extension
(CMSA-Hand) [27] between 1 and 4, inclusive (moderate joint angles were subtracted from the flexed-hand robot-
to severe hand impairment) assisted flexion joint angles to calculate the robot-assisted
• Less than 45◦ of active extension in the index finger ROM (R-A ROM). The relaxed-hand measurements were
proximal interphalangeal (PIP) joint, measured using a not used because the robot is intended to assist the par-
finger goniometer ticipants’ residual abilities as an assistive and rehabilitative
• Greater than 45◦ of passive extension in the index finger device.
PIP joint, measured using a finger goniometer 2) Grip and Pinch Strength: The participants’ grip and
• Numeric Pain Rating Scale (NPRS) score between pinch strengths were measured using a dynamometer and
0 and 4, inclusive, after active and passive finger flexion pinch gauge (JAMAR) with sensitivities of 1kg and 0.5kg.
and extension The participants’ fingers were positioned around each gauge
• No severe risk for skin breakdown under applied loads with the arm resting on the table. The researcher supported the
• No Botulinum Type A Toxin (Botox) injections in the gauge and asked the participant to grip and pinch with their
hand within the last 3 months maximum strength.
Only the PIP joint was measured to reduce the participants’ Robot-assisted grip and pinch strength was first measured
screening time commitment and because stroke survivors move while the participants were asked to keep their hand relaxed to
this finger joint the least, compared to able-bodied participants, allow the robot to deliver the grip and pinch force. Then the
while grasping [32]. participants were asked to flex their hand to provide additional
grip and pinch force. The flexed-hand grip and pinch forces
were used for the robot-assisted grip and pinch strength results.
E. Assessments Strength and ROM measurements were added to the study
1) Range of Motion, Tone and Spasticity: The stroke partici- protocol after P1.
pants were seated with their hand and arm resting on a table 3) Box and Block Test: The Box and Block Test (BBT)
at approximately elbow height. The researcher measured the is a test of participants’ capability to grasp 2cm x 2cm
920 IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 27, NO. 5, MAY 2019

TABLE II Touch (FMA-S) [36]. Extra caution was taken to check for
S TROKE PARTICIPANT D EMOGRAPHICS AND H AND F UNCTION redness and marks on the skin in these cases. No participants
reported pain while moving the joints, as assessed using the
Numeric Pain Rating Scale.

B. Unassisted Range of Motion and Strength

The researcher was able to fully straighten the index finger
MCP, PIP and DIP joints for all participants except P2 and
P5. Their MCP and DIP joints did not resist extension, but
their PIP joints resisted extension at the end range for P2 and
throughout the range of motion for P5. For all participants,
wooden blocks, lift them across a 15.2cm barrier at their the finger was able to be fully bent so the fingertip touched the
midline, and release the blocks, in one minute [29]. On palm. Active ROM was not visible in any finger (or thumb) for
average, able-bodied subjects over 75 years of age can trans- three participants (P1, P2, P4). P3 could flex and extend each
fer more than 60 blocks [34]. This test has been used to finger except the thumb and index finger PIP joint. P5 could
evaluate a previous robotic hand orthosis [30]. Participants modulate grip force to demonstrate a small active range of
who are able to perform the BBT may also be able to motion. Grip strength was not existent in four of five stroke
perform daily tasks with similar sized items, such as utensils, participants. Pinch strength was not detected in three of five
toothbrushes and handles. Participants were asked to perform stroke participants and was measured using a tripod pinch for
this task without robot assistance and with the HERO Glove in all participants except P5, whose lateral pinch strength was
both the button-press and automated mode. Participants were measured because the fingers could not be oriented properly
given up to three minutes to practice the task before being for a tripod pinch.
evaluated.
4) Chedoke Arm and Hand Activity Inventory – Water Bottle C. Robot-Assisted Range of Motion and Strength
Task: The water bottle grasp task, an ADL, was assessed using
the Chedoke Arm and Hand Activity Inventory (CAHAI) scale The HERO Glove was effective in assisting the stroke sur-
from 1 (unable to perform task) to 7 (able to perform the vivors’ motion. Each participant showed greater index finger
task independently and quickly without assistance from the extension (nearer to 0◦ ) with robot assistance than with-
unaffected hand) [35]. Participants were seated with their hand out robot assistance (increase of 97.5◦, SD 24.0, p<0.01).
resting on a table and a water bottle placed approximately A similar increase in thumb and middle finger extension was
20cm in front of their torso. Participants were instructed observed for all stroke survivors. The robot-assisted ROM was
to reach with their affected arm to grasp the water bot- larger than the active (unassisted) ROM for three of four stroke
tle, lift the water bottle and hold the water bottle while survivors (increase of 46.3◦, SD 65.5). This result was not
twisting off the lid with the opposite hand. Arm assistance significant due to the low sample size and because P3 was able
was provided by the researcher or the unaffected side if to form a tighter fist when not wearing the glove. The Percent
needed. An empty plastic water bottle was used as opposed of Motion Restored (%MR) metric is proposed in Eq. 1 for
to the coffee jar recommended for the CAHAI, because it evaluating how well the robot achieves the goal of restoring the
was safer to drop, easily accessible, of comparable diameter participants’ passive ROM. This metric compares the robot-
(76mm), and light enough to lift with a weak but active assisted ROM to the active ROM and is normalized by the
arm. difference between the passive and active ROM so that robots
can be compared evenly if the participants’ residual abilities
vary between samples.
III. S TUDY R ESULTS
(robot assited active ROM − active ROM)
A. Participants
(passive ROM − active ROM)
This study involved five acute and chronic stroke survivors = %PROM Recovered (1)
with a broad range of severe hand impairments, as presented
in Table II. The study results for ROM and task performance The HERO Glove’s assistance restored a portion of the pas-
are shown in Tables III and IV. The grip and pinch strength sive ROM for three of four participants (20.4%MR, SD 32.5).
results are shown in Supplementary Table II. The participants The largest increases in extension and %MR were observed at
ranged from CMSA-Hand level 1 (flaccid paralysis) to 3 the joints with the least active motion and tone. Participants
(able to flex but not extend the fingers). Tone and spasticity with toned (clenched) hands experienced increases in exten-
(restriction to assisted finger extension) was measured using sion at the MCP and DIP joints and the PIP joint remained
the MMAS and MTS and ranged from 0 (no increase in tone) mostly immobile. The actuator fully extended and contracted
to 2 (more marked increase in tone), with no score differences for each participant. For participants P2, P4 and P5, the robot-
between the two measures. Four of five participants showed assisted ROM was equal to the relaxed-hand robot-assisted
a reduced sense of touch in their fingers, palm and fore- ROM because these participants’ lacked active finger motion.
arm, using the Fugl-Meyer Assessment - Sensation to Light P3’s robot-assisted ROM was 15◦ larger than his robot-assisted
YURKEWICH et al.: HERO GLOVE: DEVELOPMENT AND TESTING WITH STROKE SURVIVORS 921

TABLE III
I NDEX F INGER E XTENSION AND R ANGE OF M OTION (ROM) A SSESSMENTS W ITH AND W ITHOUT R OBOT A SSISTANCE (R-A)

TABLE IV
TASK -B ASED A SSESSMENTS W ITH AND W ITHOUT R OBOT A SSISTANCE (R-A)

relaxed-hand ROM because his residual grip strength allowed D. Unassisted Box & Block Test Performance
him to further flex his MCP joint. The glove provided minor Three of five participants required support from their unaf-
obstructions to measuring joint angles; however, this was a fected hand to move their hand into the box. Four of five
more robust approach than the optical tracking system used participants could not grasp any blocks with their affected
in preliminary testing, due to the small distance between hand. P3 transferred two blocks in one minute using the little
joints, occlusion caused by other fingers and detachment of finger.
markers from the skin. The robot’s flexible structure relied
on the participants’ anatomy to restrict over-extension. The
robot extended the DIP joint past straight for three participants E. Robot-Assisted Box & Block Test Performance
because their PIP joints resisted extension while their DIP The HERO Glove enabled stroke survivors to incorporate
joints were flaccid. No participants reported pain from this their affected hand into the BBT and water bottle task,
motion. Two participants reported mild pain, NPRS 1-2 (out as shown in Fig. 2, Fig. 3 and Table IV. The HERO Glove
of 10), on the dorsal side of the index finger’s proximal enabled each participant to create space between the fingers
phalange after more than thirty minutes of use that resolved and the thumb during extension and then touch their index
with rest. finger to their thumb, to create a tripod pinch for four partici-
The robot’s assistance did not enhance the participants’ pants and a lateral pinch for P5. This assistance improved each
grip or pinch strength. The flexible structure of the robot participant’s performance in grasping and transferring blocks
caused the fingers and thumb to divert from the reaction with the HERO Glove (2.8 block increase, SD 1.3, p<0.01).
force of the dynamometer and pinch gauge. Instead of apply- Since each participant had an inability to lift their affected
ing force through the MCP joint, the PIP and DIP joints arm for one minute the researcher supported and positioned
curled around the dynamometer and deflected off the pinch the forearm. The glove did not fully extend toned hands so the
gauge. researcher operated the physical button, knowing when the
922 IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 27, NO. 5, MAY 2019

Fig. 2. The HERO Glove enabled stroke participants to transfer blocks during the Box and Block Test. The left figure demonstrates P1 attempting
unsuccessfully to grasp a block with the affected hand, while supporting the arm with the other hand. The two figures on the right demonstrate
P1 completing this block grasp and transfer task independently with the HERO Glove providing extension and flexion assistance automatically.

Fig. 3. Five stroke survivors with severe hand impairments evaluated the HERO Glove’s assistive capabiltities. The top images show the
hand impairments of P1 through P5, respectively. The bottom images show how the HERO Glove was worn and operated by P1 through P5,
respectively, to perform tasks independently. The HERO Glove’s assistance enabled P3 and P5 to extend their fingers and thumb around the
water bottle. The HERO Glove’s assistance extended and flexed P1, P2 and P4’s fully paralyzed hands to grasp the water bottle and blocks
independently.

hand was best-oriented for the grasp. Then each participant F. Unassisted Water Bottle Task Performance
trialed the manual and automated modes and supported their No participants could complete the water bottle ADL task
forearm with their unaffected hand. Each participant was able without the unaffected hand supporting the affected hand. Only
to control the robot’s assistance in both the button-press and P3 and P5 had some capacity to reach with their affected arm,
automated modes. P2, P4 and P5 did not trial the automated although limited due to weakness. P5 could not extend the
mode independently during the BBT because they were unable fingers enough to press the water bottle into the affected hand.
to move their affected arm with the accuracy required position P1, P2, and P4 did not have the grip strength required to hold
their fingers around the blocks. P1 and P3 were able to use the water bottle while lifting or twisting the lid. P3 required
the HERO Glove in the automated mode to each grasp and assistance from the other hand to stretch the affected fingers
transfer two blocks in one minute independently. P1 required and then quickly press the water bottle into the affected hand.
the automated mode because it was difficult to press the P3 could then hold and lift the water bottle while removing
button while supporting the full weight of his flaccid arm. the lid without assistance.
P3 transferred five blocks independently in the button-press
mode and preferred this mode to the automatic mode for its
reliability and switching speed. P5 transferred three blocks G. Robot-Assisted Water Bottle Task Performance
in the automated mode, with the researcher supporting her The water bottle task demonstrated the HERO Glove’s
forearm. The automated mode functioned perfectly for P1, assistive capabilities and areas for design improvement. P1, P2,
P3 and P5, as all seven grasps and seven extensions required to and P5 showed improved performance, as assessed using the
transfer the seven blocks occurred when desired and without CAHAI scale, with the HERO Glove enabling them to extend
added delay or early release of blocks (0% false positives, 0% their fingers and place the water bottle in their hand. The glove
false negatives). All participants were experiencing global and enabled them to hold the water bottle during lifting and lid
muscular fatigue by one hour into the study so the BBT trials twisting. The CAHAI scores for P3 and P4 did not change
were not repeated. because the glove did not provide enough thumb extension
YURKEWICH et al.: HERO GLOVE: DEVELOPMENT AND TESTING WITH STROKE SURVIVORS 923

for P3 to complete the grasp unassisted or enough force for The HERO Glove’s assistive capacity was evaluated with five
P4 to hold the water bottle while twisting off the lid. The stroke survivors with severe hand impairment and provides
water bottle task was trained using a hand-over-hand technique evidence of its efficacy by demonstrating a:
for less than three minutes and was assessed while the stroke • Statistically significant increase in index finger extension
survivors performed the task independently in the button-press for stroke survivors with flaccid and toned hands.
mode. P3 successfully used the automated mode to trigger • A statistically significant improvement in performance on
extension and flexion to grasp, lift, lower and release the water a functional task, the BBT, and improvement for most
bottle. The lid was not removed because lifting then stopping participants in ROM and on an ADL, the water bottle
caused the glove to release the water bottle. The automated task.
mode was not tested with the other participants due to arm This work also provides design guidance for further wear-
fatigue from the prior assessments. The HERO Glove’s weight able hand robot development through the requirements sug-
did not affect P3 or P5’s ability to reach and lift their arms gested by therapists and stroke survivors and observations on
while holding the water bottle. how well five stroke survivors with severe hand impairment
used the HERO Glove in manual and automated control
H. Usability Observations With the HERO Glove modes, as summarized in Table I. In its current development
The stroke participants and occupational therapists were stage, the HERO Glove can help stroke survivors with specific
informally questioned about the glove’s usability after hand impairments perform daily tasks, such as holding bowls,
the trials. Table I summarizes how well the HERO Glove containers and pans that require object stabilization but not
met their requested specifications. They expressed satisfac- lifting. The HERO Glove should provide greater grip strength
tion with the HERO Glove’s motivations as an assistive and through the fingers and thumb to enable individuals with
rehabilitative device for performing daily tasks more easily weak grip strength to independently perform daily activities
and independently and reintegrating the affected hand. Their safely.
satisfaction with its portability, light weight such that it did
not affect arm motion or fatigue, ease of donning, set up
A. Impact of Mechanical Design on Task Performance
and use and potential affordability provided justification for
the untethered design. They commented that the grip strength The HERO Glove accomplished its main objective of
should be improved, an arm support should be available, increasing finger extension so stroke survivors with severe
and the construction should be more comfortable, robust and hand impairments could grasp daily objects. The cable tie
aesthetically-pleasing for the stroke survivors to use the HERO tendons applied a strong force to extend high-tone fingers,
Glove during daily tasks at home. which has proven to be difficult for previous robots [20], [21].
The glove extended the DIP and MCP joints fully for each
IV. D ISCUSSION participant; however, the stroke survivors with high finger tone
Robotic hand orthoses have the potential to enable stroke also desired full restoration of extension at their stiffest joint,
survivors to generate larger motions and stronger forces. the PIP joint. The challenge for increasing this extension is in
This can enable stroke survivors to more usefully incorporate creating a mounting point on the glove that mounts the single
their affected hand into activities of daily living that would actuator as rigidly as a wrist brace with distributed actuators,
otherwise require compensatory strategies and caregiver sup- as in Fischer et al. [12]. In addition, the spacing between cable
port. We iteratively designed a novel robotic hand orthosis guides should be reduced to block hyperextension and further
and control strategy with occupational therapists and stroke distribute pressure, as in [12] and Rose and O’Malley [37].
survivors based on their specified requirements. Key novel Once the glove was removed, the hand was less toned and
features of the HERO Glove are: more extended, which may motivate spontaneous unassisted
• The robot is untethered and fully contained on the hand, hand use.
including the mechanism, actuator, electronics and bat- All participants with flaccid hands and one participant with
tery. This minimizes the number of donning steps and a toned hand could not produce a strong grip force. With a
makes the device wireless and convenient to use when passive dynamic orthosis these participants would not have
sitting, standing and transferring. been able to overcome the extension bias force, leading to
• The buckling-resistant dorsal cable ties are coupled to a poorer task performance [38]–[40]. In contrast, the HERO
single motor to provide strong extension assistance and Glove’s actuator contracts to release the extension force and
some flexion assistance. This enables the HERO Glove assist the fingers to flex, increasing their ROM. In addition,
to be more affordable for stroke survivors, have an open- quantifying how well the assistance increases survivors’ finger
palm to ease donning on a flexed hand and possess extension and active ROM at a joint level provides a more
the lowest overall weight among wearable robotic hand specific benchmark for future hand robot evaluations than the
orthoses to minimize arm fatigue. FMA-UE [20]. Stronger grip force is required to improve
• The use of an IMU to measure the user’s arm and wrist task performance. Methods for integrating flexor tendons and
motion and use this signal to trigger robotic assistance. donning the HERO Glove independently should be inves-
This enables stroke survivors with severe hand impair- tigated, such as routing the tendons dorsally or attaching
ment to control the HERO Glove without their unaffected the tendons with buckles or ratchets once the fingers are
hand. extended [37], [41], [42].
924 IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 27, NO. 5, MAY 2019

The HERO Glove is the first wearable hand robot to Given the low false negative and positive rates demonstrated
show that its assistive capacity enhances BBT performance in the BBT, the low cost and size of IMUs and the infancy of
for stroke survivors. The HERO Glove is untethered, which IMU-triggered orthoses, there is an opportunity and motivation
differentiates it from previous wearable hand robots that to improve algorithm reliability. Insight can be gained from
improved block grasping for spinal cord injury and mus- Bennett and Goldfarb [46], where a non-synergistic move-
cular dystrophy, but required wheelchair-mounted actuation ment (shoulder abduction) was used to intuitively control a
units [22], [25], [43]. This study extends current evidence that prosthetic wrist’s pronation velocity. Participants could also
untethered wearable hand robots can effectively assist stroke be trained to cancel misfires by shaking their hand, similar to
survivors’ cylindrical grasp [20], [44], by demonstrating that how electrooculography has been used [25]. Machine learning
three stroke survivors were only able to perform the water approaches could be used to fuse IMU data with force, vision,
bottle task with the HERO Glove’s assistance. Performance voice, EMG and EEG data to improve reliability and add
could be improved by further assisting finger extension, grip dexterity and grasp force modulation for higher degree-of-
strength and thumb abduction and opposition. The actuator freedom exoskeletons.
could be relocated closer to the thumb to better assist its A main constraint for this design was the need to keep
motion, but this may obstruct wrist motion and affect the the material cost affordable so the device could more
glove’s aesthetics [12]. Additional studies are required to likely be accessed by stroke survivors without comprehensive
investigate how well the HERO Glove assists stroke survivors health insurance coverage. The design incorporates only one
in a variety of activities of daily living, using the CAHAI and actuator, minimal sensors and low-cost components to keep
Toronto Rehabilitation Institute Hand Function Test [22], [25]. the cost within the range specified by stroke survivors [23].
To perform these activities independently, upper-arm neu- Do-It-Yourself communities could assemble the HERO Glove
roprostheses like exoskeletons, gravity supports and neuro- themselves to minimize manufacturing costs, as is currently
muscular stimulators may be necessary because each stroke managed with elbow exoskeleton kits [47]. Personalizing the
participant showed significant weakness in shoulder flexion glove to the user’s hand size and swelling would enhance
and elbow extension and fatigued quickly. the glove’s ease of donning, comfort and assistive capacity.
An actuator with a lower gear ratio that is faster and back-
drivable could also be selected if the user does not need a
B. Usability Perspectives of Therapists and strong extension force. In addition, the IMU control thresholds
Stroke Survivors could be tuned to the individual. Given the diversity in digital
The overarching goal of this iterative design process was to literacy, lifestyle and upper extremity function among stroke
create a wearable hand robot that met therapists’ and stroke survivors, multiple designs and adaptive control schemes may
survivors’ requirements so the robot would be easy to integrate be required to serve specific subsets of the stroke population.
into therapy practice and daily routines. To meet this need,
we created a portable, easy to use and affordable device that
enables stroke survivors with low-functioning hands to practice C. Study Limitations
higher level tasks that are more similar to their daily tasks. This paper presents the HERO Glove design and an obser-
The therapists were interested in using the HERO Glove to vational clinical pilot study evaluating its assistive efficacy.
practice more engaging real-world activities with their clients. The diversity of experience within our design team and the
They suggested that the glove could help clients adhere to the rolling recruitment method allowed us to understand the
forced-use component within programs like constraint-induced complexities of two stroke survivors’ hand impairment and
movement therapy [45]. The HERO Glove is currently suitable revise the glove’s structure, form, fit and control to meet future
for stroke survivors, CMSA-Hand <4, that require greater participants’ hand assistance needs. Limitations of this study
finger extension and flexion in order to work towards their design are that the statistical power is low and the solution
therapy goals in the clinic. Before the HERO Glove is ready for may not be effective for stroke survivors with only mild to
home use, further design is required to block hyperextension, moderate hand impairment.
distribute pressure and replace glued areas with bolts and In future study designs it would be useful to have the
sewn on enclosures that protect the wires and mechanism from same participants return to test device iterations to validate
impact, snagging and continuous wear. In addition, the ther- the usefulness of the modifications. This would also allow us
apists requested improved assistive capabilities, especially for to perform repeat trials and quantitative experiments on ROM
grip strength and arm support, so they would have more of the thumb and other fingers, muscle activity and holding
flexibility in selecting tasks that align with their clients’ goals. force before the participant experiences fatigue. Collecting
The button-press mode was reliable and easy to understand the IMU data would also be useful for quantifying arm
with little training. The therapists and stroke survivors voiced motion; however, firmware updates are required to transmit
that the automated mode was important because it kept the the data via Bluetooth without delaying the automated control
unaffected hand free and could motivate spontaneous arm use. mode. Usability feedback from our participants and stroke
The current algorithm is useful for pick-and-place exercises working group provided an understanding of their experi-
but the algorithm needs to be as reliable as the button-press ences with the HERO Glove, but usability questionnaires,
mode to give users confidence in spontaneously using their such as the Psychosocial Impact of Assistive Devices Scale
arm during daily use without dropping breakable objects. (PIADS) [40] or Usefulness, Satisfaction and Ease of Use
YURKEWICH et al.: HERO GLOVE: DEVELOPMENT AND TESTING WITH STROKE SURVIVORS 925

questionnaire [23], and semi-structured interviews are required [11] E. Susanto, R. K. Tong, C. Ockenfeld, and N. S. Ho, “Efficacy of robot-
to provide structured design guidance. assisted fingers training in chronic stroke survivors: A pilot randomized-
controlled trial,” J. Neuroeng. Rehabil., vol. 12, no. 1, p. 42, Dec. 2015.
[12] H. C. Fischer et al., “Use of a portable assistive glove to facili-
V. F UTURE W ORK tate rehabilitation in stroke survivors with severe hand impairment,”
IEEE Trans. Neural Syst. Rehabil. Eng., vol. 24, no. 3, pp. 344–351,
Key lessons learned through this study were that therapists Mar. 2016.
and stroke survivors can play a vital role in tailoring the [13] S. M. Nijenhuis, G. B. Prange-Lasonder, A. H. Stienen, J. S. Rietman,
and J. H. Buurke, “Effects of training with a passive hand orthosis and
device’s usability and stroke survivors are interested in using games at home in chronic stroke: A pilot randomised controlled trial,”
wearable hand robots to assist their affected hand to perform Clin. Rehabil., vol. 31, no. 2, pp. 207–216, Feb. 2016.
daily tasks independently. There is a strong need to design [14] F. Vanoglio et al., “Feasibility and efficacy of a robotic device
for hand rehabilitation in hemiplegic stroke patients: A randomized
and evaluate these robots with a greater number and variety pilot controlled study,” Clin. Rehabil., vol. 31, no. 3, pp. 351–360,
of people with neurologically-affected hands. This will help Mar. 2016.
refine design specifications to improve their assistive and [15] J. M. Veerbeek, A. C. Langbroek-Amersfoort, E. E. H. van Wegen,
C. G. M. Meskers, and G. Kwakkel, “Effects of robot-assisted therapy
rehabilitative efficacy, as currently no robots can be used for the upper limb after stroke: A systematic review and meta-analysis,”
independently or enable affected hands to perform activities Neurorehabil. Neural Repair, vol. 31, no. 2, pp. 107–121, 2016.
of daily living as functionally or quickly as unaffected hands. [16] J. Mehrholz, M. Pohl, T. Platz, J. Kugler, and B. Elsner, “Electromechan-
ical and robot-assisted arm training for improving activities of daily
Once the HERO Glove is sufficiently advanced to satisfy living, arm function, and arm muscle strength after stroke ( Review ),”
stroke survivors ADL needs, we will be able to study how well Cochrane Database Syst. Rev., to be published.
its assistance can be integrated into a home therapy program to [17] S. J. Page, V. Hill, and S. White, “Portable upper extremity robotics is
as efficacious as upper extremity rehabilitative therapy: A randomized
promote neuromuscular recovery. Through the addition of data controlled pilot trial,” Clin. Rehabil., vol. 27, no. 6, pp. 494–503,
storage and a mobile application we will be able to monitor Jun. 2013.
their program adherence and recovery and provide feedback [18] G. J. Kim, L. Rivera, and J. Stein, “Combined clinic-home
approach for upper limb robotic therapy after stroke: A pilot
to therapists to learn their clients’ behavior and adapt their study,” Arch. Phys. Med. Rehabil., vol. 96, no. 12, pp. 2243–2248,
therapy program to meet their changing capacities and needs. Dec. 2015.
[19] S. M. Nijenhuis et al., “Feasibility study into self-administered training
at home using an arm and hand device with motivational gaming
ACKNOWLEDGMENTS environment in chronic stroke,” J. Neuroeng. Rehabil., vol. 12, no. 1,
p. 89, Dec. 2015.
Special thanks to the members of the Intelligent Assistive [20] H. T. Peters, L. Otr, S. J. Page, L. Otr, A. Persch, and L. Otr,
Technology and Systems Lab (IATSL), and Daniel Rossos, “Giving them a hand: Wearing a myoelectric elbow-wrist-hand orthosis
Sylvia Haycock, Cynthia Ho, Marge Coahran and Michael reduces upper extremity impairment in chronic stroke,” Arch. Phys. Med.
Rehabil., vol. 98, pp. 1821–1827, Sep. 2017.
Jaunkalns for their recommendations and feedback. [21] S. Park, L. Bishop, T. Post, Y. Xiao, J. Stein, and M. Ciocarlie, “On the
R EFERENCES feasibility of wearable exotendon networks for whole-hand movement
patterns in stroke patients,” in Proc. IEEE Int. Conf. Robot. Autom.,
[1] A. S. Go, D. Mozaffarian, V. L. Roger, E. J. Benjamin, J. D. Berry, and May 2016, pp. 3729–3735.
S. Franco, “Heart disease and stroke statistics-2014 update: A report [22] L. Cappello et al., “Assisting hand function after spinal cord injury with
from the american heart association,” Circulat, vol. 129, no. 3, p. e28, a fabric-based soft robotic glove,” J. Neuroeng. Rehabil., vol. 15, no. 1,
2014. pp. 1–10, Dec. 2018.
[2] G. E. Gresham, T. E. Fitzpatrick, P. A. Wolf, P. M. McNamara, [23] H. K. Yap, J. H. Lim, F. Nasrallah, and C. H. Yeow, “Design and
W. B. Kannel, and T. R. Dawber, “Residual disability in survivors of preliminary feasibility study of a soft robotic glove for hand function
stroke-the framingham study,” New England J. Med., vol. 293, no. 19, assistance in stroke survivors,” Front. Neurosci., vol. 11, pp. 1–14,
pp. 954–956, Nov. 1975. Oct. 2017.
[3] A. K. Welmer, L. W. Holmqvist, and D. K. Sommerfeld, “Limited [24] K. O. Thielbar et al., “Benefits of Using a Voice and EMG-Driven
fine hand use after stroke and its association with other disabilities,” Actuated Glove to Support Occupational Therapy for Stroke Survivors,”
J. Rehabil. Med., vol. 40, no. 8, pp. 603–608, Sep. 2008. IEEE Trans. Neural Syst. Rehabil. Eng., vol. 25, no. 3, pp. 297–305,
[4] C. J. Winstein et al., “A randomized controlled comparison of upper- Mar. 2017.
extremity rehabilitation strategies in acute stroke: A pilot study of [25] S. R. Soekadar et al., “Hybrid EEG / EOG-based brain / neural
immediate and long-term outcomes,” Arch. Phys. Med. Rehabil., vol. 85, hand exoskeleton restores fully independent daily living activities after
pp. 620–628, Apr. 2004. quadriplegia,” Sci. Robot., vol. 1, no. 1, pp. 32–96, 2016.
[5] S. M. Lai, S. Studenski, P. W. Duncan, and S. Perera, “Persisting [26] C. J. Nycz, M. A. Delph, and G. S. Fischer, “Modeling and design of
consequences of stroke measured by the stroke impact scale,” Stroke, a tendon actuated soft robotic exoskeleton for hemiparetic upper limb
vol. 33, no. 7, pp. 1840–1844, Jul. 2002. rehabilitation,” in Proc. Int. Conf. IEEE Eng. Med. Biol. Soc. EMBS,
[6] C. G. Burgar et al., “Robot-assisted upper-limb therapy in acute Aug. 2015, pp. 3889–3892.
rehabilitation setting following stroke: Department of veterans affairs [27] P. Miller et al., “Chedoke-McMaster stroke assessment—Development,
multisite clinical trial,” J. Rehabil. Res. Dev., vol. 48, no. 4, pp. 445–458, validation and administration manual,” McMaster Univ. Hamilton
2011. Health Sci., Hamilton, ON, Canada, Tech. Rep. 2nd Edition, Oct. 2008.
[7] R. Viana and R. Teasell, “Barriers to the implementation of constraint- [Online]. Available: https://www.sralab.org/sites/default/files/2017-
induced movement therapy into practice,” Topics Stroke Rehabil., vol. 19, 07/CMSA%20Manual%20and%20Score%20Form.pdf
no. 2, pp. 104–114, Mar. 2012. [28] N. N. Ansari, S. Naghdi, M. Mashayekhi, S. Hasson, Z. Fakhari, and
[8] G. B. Prange et al., “User requirements for assistance of the supporting S. Jalaie, “Intra-rater reliability of the modified modified ashworth
hand in bimanual daily activities via a robotic glove for severely affected scale (MMAS) in the assessment of upper-limb muscle spasticity,”
stroke patients,” in Proc. IEEE Int. Conf. Rehabil. Robot., Aug. 2015, NeuroRehabilitation, vol. 31, no. 2, pp. 215–222, Jan. 2012.
pp. 357–361. [29] V. Mathiowetz, G. Volland, N. Kashman, and K. Weber, “Adult norms for
[9] S. Ates, I. Mora-Moreno, M. Wessels, and A. H. A. Stienen, “Combined the box and block test of manual dexterity,” Arch. Phys. Med. Rehabil.,
active wrist and hand orthosis for home use: Lessons learned,” in Proc. vol. 39, no. 6, pp. 386–391, 1985.
IEEE Int. Conf. Rehabil. Robot., Aug. 2015, pp. 398–403. [30] P. Polygerinos, K. C. Galloway, S. Sanan, M. Herman, and C. J. Walsh,
[10] R. A. Bos et al., “A structured overview of trends and technologies used “EMG controlled soft robotic glove for assistance during activities of
in dynamic hand orthoses,” J. Neuroeng. Rehabil., vol. 13, no. 1, p. 62, daily living,” in Proc. IEEE Int. Conf. Rehabil. Robot., Aug. 2015,
Dec. 2016. pp. 55–60.
926 IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 27, NO. 5, MAY 2019

[31] M. Gandolla et al., “Technical validation of an integrated robotic hand [39] B. T. Iwamuro, H. C. Fischer, and D. G. Kamper, “A pilot study to
rehabilitation device: Finger independent movement, EMG control, and assess use of passive extension bias to facilitate finger movement for
EEG-based biofeedback,” in Proc. IEEE 2nd Int. Forum Res. Technol. repetitive task practice after stroke,” Top. Stroke Rehabil., vol. 18, no. 4,
Soc. Ind., Sep. 2016, pp. 1–4. pp. 308–315, Jul. 2011.
[32] P. Raghavan, M. Santello, A. M. Gordon, and J. W. Krakauer, “Com- [40] L. Andriske, D. Verikios, and D. Hitch, “Patient and ther-
pensatory motor control after stroke: An alternative joint strategy for apist experiences of the SaeboFlex: A pilot study,” Occupa-
object-dependent shaping of hand posture,” J. Neurophysiol., vol. 103, tional Therapy Int., vol. 2017, Jan. 2017. [Online]. Available:
no. 6, pp. 3034–3043, 2010. https://www.ncbi.nlm.nih.gov/pubmed/29097972
[33] R. N. Boyd and H. K. Graham, “Objective measurement of clinical [41] D. H. Kim, S. Heo, and H. Park, “Biomimetic finger extension mecha-
findings in the use of botulinum toxin type A for the management nism for soft wearable hand rehabilitation devices,” in Proc. Int. Conf.
of children with cerebral palsy,” Eur. J. Neurol., vol. 6, no. S4, Rehabil. Robot., Feb. 2017, pp. 1326–1330.
pp. s23–s35, 1999. [Online]. Available: https://www.researchgate.net/file. [42] D. Popov, I. Gaponov, and J. H. Ryu, “Portable exoskeleton glove
PostFileLoader.html?id=54cbf0bfd039b1664e8b4677&assetKey=AS% with soft structure for hand assistance in activities of daily liv-
3A273685905117184%401442263212921 ing,” IEEE/ASME Trans. Mechatron., vol. 22, no. 2, pp. 865–875,
[34] J. Dewosiers, C. Ot, G. Bravo, and R. Hibert, “Validation of the box and Apr. 2017.
block test as a measure of dexterity of elderly people: Reliability, valid-
[43] P. Polygerinos, K. C. Galloway, E. Savage, M. Herman, K. O’Donnell,
ity, and norms studies,” Arch. Phys. Med. Rehabil., vol. 75, pp. 751–755,
and C. J. Walsh, “Soft robotic glove for hand rehabilitation and task
Jul. 1994.
specific training,” in Proc. IEEE Int. Conf. Robot. Autom., May 2015,
[35] S. Barreca et al., “Development of the chedoke arm and hand activity
pp. 2913–2919.
inventory: Theoretical constructs, item generation, and selection,” Top.
Stroke Rehabil., vol. 11, no. 4, pp. 31–42, 2004. [Online]. Available: [44] B. W. Gasser, D. A. Bennett, C. M. Durrough, and M. Goldfarb, “Design
http://www.cahai.ca/layout/content/CAHAI-Manual-English.pdf and preliminary assessment of vanderbilt hand exoskeleton,” in Proc. Int.
[36] J. Sanford, J. Moreland, L. R. Swanson, P. W. Stratford, and C. Gowland, Conf. Rehabil. Robot., Feb. 2017, pp. 1537–1542.
“Reliability of the Fugl-meyer assessment for testing motor performance [45] M. H. Bowman et al., “A treatment for a chronic stroke patient with a
in patients following stroke,” Phys. Ther., vol. 73, no. 7, pp. 445–447, plegic hand combining CI therapy with conventional rehabilitation pro-
1993. cedures: Case report,” NeuroRehabilitation, vol. 21, no. 2, pp. 76–167,
[37] C. G. Rose and M. K. O’Malley, “Hybrid rigid-soft hand exoskeleton 2006.
to assist functional dexterity,” IEEE Robot. Autom. Lett., vol. 4, no. 1, [46] D. A. Bennett and M. Goldfarb, “IMU-based wrist rotation control of
pp. 73–80, Jan. 2019. a transradial myoelectric prosthesis,” IEEE Trans. Neural Syst. Rehabil.
[38] S. M. Nijenhuis, G. B. Prange, A. H. A. Stienentt, J. H. Buurke, and Eng., vol. 26, no. 2, pp. 419–427, Feb. 2018.
J. S. Rietman, “Direct effect of a dynamic wrist and hand orthosis on [47] M. Scherzinger and G. Garcia, Entwicklungeines Ellenbogens-
reach and grasp kinematics in chronic stroke,” in Proc. IEEE Int. Conf. Exoskeletts Development of an elbow’s Exoskeleton Handbook and
Rehabil. Robot., Aug. 2015, pp. 404–409. Tutorial. Nanjing, China: ELEX, 2017.