Journal of Rehabilitation Medicine 51-4inkOmslag | Page 67

UE performance with a soft-robotic glove in elderly patients
Data analysis
The recorded movement data were analysed using VICON
Nexus 1.8.2 and transferred to MATLAB software (R2015a,
MathWorks Inc., Natick, MA, USA) for further custom analysis.
All position data were low-pass filtered with a second-order
Butterworth filter of 20 Hz with zero phase shift.
Segmentation
The recorded data started at the first, and ended at the last mo-
ment that the tangential velocity of the hand exceeded 2% of the
maximum velocity of the hand based on the second metacarpal
marker (24). From that recording, 4 task phases were identified
as described below and presented in Fig. 3.
The start of the grasping phase was identified through either
the position of the cylindrical object relative to the hand or the
velocity profile of the object within the timeframe between index
2 and index 3 (see Fig. 3). The indices were chosen at 15% of
the maximal velocity of the hand to facilitate the determination
of the movement phases. The position of the object relative
to the hand reaches a minimum when an attempt is made to
grasp. In addition, when the object’s velocity is larger (mean
+2 times standard deviation (SD)) for the first time than when
the cylindrical object stood still, it is likely that the cylindrical
object is moved by an external source. The start was set at the
lowest frame number of the 2 options. The end of grasping was
defined as the last frame number of:
• the difference in velocity of the object relative to the hand
marker was smaller than the mean value minus 2 times the SD
of the velocity of the object, as measured when that object was
not moved (Fig. 5). During the reach-with-object phase, the
hand and cylinder were expected to have the same movement
pattern. The difference in velocity of the hand and cylinder
will therefore be minimal; or:
• the latest minimum of the vertical position of the object in
time between index 2 and index 3 (Fig. 3). Before the object
was lifted in the vertical direction, a minimum was seen in
the vertical position profile of the object; or:
• the combined XYZ-position of the object is larger than 2
times the SD plus the mean combined XYZ value of the object
when not moved, and if 15 frames later the XYZ position of
the object is more than 12 times the SD plus the mean of the
object when not moved. If grasped, the object’s movement
was not necessarily in the vertical direction. In the case that
the object was moved due to touching instead of measuring
the end of grasping, the condition that the object had to be
moved substantially (12 times the SD + mean) some frames
later was built in.
The start of the reach-with-object phase coincided with the
end of the grasping phase. The end of this phase was set at the
frame number when the object touched the platform. This event
was chosen, because the object was always lifted higher than
the height of the platform, after which the impact of the object
with the platform caused a minimum in the vertical position
of the object.
Releasing started at the end of the reach-with-object phase
and ended with the last frame number of either:
• the first time that the difference in position of the object and
hand marker exceeded the minimal distance, as determined
in the grasping phase, between object and hand plus 0.15
times the SD; or:
• the last time that the velocity of the object exceeded the mean
velocity plus 2 times SD of the object as measured in rest.
301
The start of the final reach coincided with the end of the
releasing phase and the end of this phase occurred at the first
time that the tangential velocity of the hand was smaller than
2% of the maximum speed of the hand (24).
The grasping and releasing phases were manually checked
by comparing the frame numbers with the visualized VICON
data. If the algorithm and manual check deviated more than
0.05 s, parameters associated with that grasp or release phase
were removed from further analysis.
Outcome measures
The primary outcome measure, total movement duration, was
calculated from the initiation of grasping until the end of the
final reach.
Secondly, the influence of the glove on time needed to execute
each of the 4 phases was calculated in absolute and relative
(percentage of total movement time) duration. The time prior to
grasping was used to calculate the mean and SD of the motionless
cylindrical object. Outcome measures comparable to previous
studies performing kinematic analyses of a reach-and-grasp task
were calculated to assess the influence of the ironHand glove on
movement execution (25–27). Smoothness of the movement,
expressed as the number of movement units (NMUs) (28),
was calculated over the entire movement. Local minima and
maxima in the tangential velocity profile of the marker on the
second metacarpal head were searched for the determination of a
movement unit. The difference between a consecutive minimum
and maximum with an amplitude of 20 mm/s or more indicated a
velocity peak that corresponds to the smoothness and efficiency
of movement (28). A movement unit was identified when the time
between 2 consecutive peaks exceeded 150 ms (24). Maximal
trunk displacement (TD) was defined as the maximal 3D displa-
cement of the trunk marker during the task compared with the
initial position in rest. Maximum speed during the reach-with-
object phase was calculated from the tangential velocity profile
of the hand, based on the marker positioned at the head of the
second metacarpal. Maximal hand opening prior to grasping
was calculated as the maximal distance between the thumb and
middle finger marker. The elbow angle during the entire task
was calculated from the angle between the vector of the upper
and lower arm. Joint excursion of the elbow was determined by
subtraction of the smallest angle from the largest angle between
those vectors. Maximum elbow extension angle was measured,
and determined as the largest angle between the upper and lower
arm. Excursion of the wrist was calculated by subtracting the
smallest angle from the largest angle between the forearm and
hand in flexion and extension direction.
Statistical analysis
Individual values across participants were averaged per task,
glove condition and weight of the cylindrical object. Values
per parameter were reported as median with interquartile range
Table I. Demographic characteristics
Characteristics
Sex, female/male, n
Age, years , median (IQR)
Most-affected body side, right/left/both a , n
Dominant side, right/left, n
Handgrip strength, kg b , median (IQR)
n  = 8
8/0
65.5 (62.3–76.5)
5/1/2
7/1
11.5 (8.0–18.0)
a
The glove was worn on the dominant hand if both sides were most-affected.
One missing value.
b
J Rehabil Med 51, 2019