The effect of an assist-as-needed algorithm applied by a robotic gait trainer on walking in stroke and SCI patients

The last decade, there is an increasing interest in the use of robotic gait
trainers during the rehabilitation of spinal cord injured patients and patients
who suffered from a stroke. Current robotic gait trainers have not been proven
to be more effective than other gait training. These robotic gait trainers are
position-controlled and have limited Degrees of Freedom. Treatment outcome
could be optimized by promoting active participation of the patient. A more
active approach is the assist-as-needed algorithm as implemented in LOPES II.
LOPES II is an impedance controller, which is able to adjust joint stiffness to
the capabilities of the patient from high (robot-in-charge) to zero
(patient-in-charge). In addition, LOPES II is a selective task controller. It
is capable of only supporting selected phases of the gait cycle that are
affected. Because of the increase in Degrees of Freedom compensatory movements
are allowed by LOPES. Besides being used as a gait trainer, LOPES II can be
used to measure joint movements and spatio-temporal parameters during gait.
The implementation of an assist-as-needed algorithm in human is quite new.
Little is known about the underlying mechanism of this algorithm. This study
will focus on the different effects of the assist-as-needed algorithm on
walking compared to the position-controlled algorithm. In addition to the
manual assist-as-needed (mAAN) which is mentioned above, there is also a
robotic assist-as-needed (rAAN) algorithm. The rAAN algorithm automatically
adjusts the amount of support whereas a physiotherapist determines the amount
of support for the mAAN. This study will also focus on the differences between
these AAN algorithms.

Primary objectives:
1. What is the effect of walking on a treadmill compared to walking in a
patient-in-charge mode on kinematics in healthy subjects?
2. What is the effect of walking on a treadmill compared to walking in a
patient-in-charge mode on kinematics in chronic SCI patients and stroke
patients?
3. What is the short term effect of walking in an mAAN algorithm compared to a
PC algorithm on kinematics in (sub)acute SCI patients and stroke
patients?
4. What is the effect of training with the mAAN algorithm compared to training
with a PC algorithm on kinematics in (sub)acute SCI patients and
stroke patients?

Secondary objectives:
5. What is the difference in EMG patterns while walking on a treadmill
compared to walking in a patient-in-charge mode in healthy subjects?
6. What is the difference in EMG patterns while walking on a treadmill
compared to walking in a patient-in-charge mode in chronic SCI patients
and stroke patients?
7. What is the short term effect of walking in an mAAN algorithm compared to
a PC algorithm on EMG patterns in (sub)acut SCI patients and stroke
patients?
8. What is the short term effect of walking in an mAAN algorithm compared to
a PC algorithm on energy consumption in (sub)acute SCI patients
and stroke patients?
9. What is the effect of training with the mAAN algorithm compared to
training with a PC algorithm on muscle activation patterns in (sub)acute SCI
patients and stroke patients?
10. What is the effect of training with the mAAN algorithm compared to training
with a PC algorithm on walking ability in (sub)acute SCI patients and
stroke patients?
11. What is the effect of training with the mAAN algorithm compared to training
with a PC algorithm on energy consumption in (sub)acute SCI
patients and stroke patients?
12. What is the short term effect of rAAN on walking ability and EMG activity
compared to an mAAN algorithm?
13. What is the difference in the amount of support given with an rAAN
algorithm compared to an mAAN algorithm?
14. What is the difference in patients experience of an rAAN algorithm compared
to an mAAN algorithm?

The study consists of four parts.
1. The first part is a cross-sectional trial in which healthy subjects, chronic
SCI subjects and chonic stroke subjects are asked to walk on a treadmill
and in LOPES II (patient-in-charge mode). The goal is to see if the
patient-in-charge mode of LOPES II interact with the patient. This
measurement is necassary for using the patient-in-charge mode is a
measurment tool for quality of walking in the following parts. Kinematics,
spatio-temporal parameters an surface EMG will be measured.
2. The second part is a cross-sectional trial with SCI and stroke subjects.
Subjects will walk with both the assist-as-needed (mAAN) and
position-controlled algorithm during minimal 20 strides. The order will
randomly be assigned. LOPES II starts with a baseline measurment in the
patient-in-charge mode, following with one support algorithm and pass over to
the patient-in-charge mode for measuring the direct effect on kinematics and
spatio-temporal parameters. EMG patterns and energy consumption will also be
measured.
3. The third part is a blinded randomised clinincal trial with a training
period of 6 weeks (3 times/week) for SCI and stroke subjects during their
inpatient rehabilitation program. Subjects will be randomly assigned for
training with the assist-as-needed algorithm (mAAN) or training with the
position-control algorithm. During the training heart rate and the Borg Scale
will be recorded. Pre-, between, post and follow up measurements will be done,
measuring walking quality, ability and muscle activation patterns.
Stratification will take place for diagnosis and walking speed.
4. The fourth part is a cross-sectional trial with (sub)acute SCI and stroke
subjects. Subjects will be asked to walk maximal 5 minutes with the manual
assist-as-needed (mAAN) as well as the robotic assist-as-needed (rAAN)
algorithm. The order of the trials is randomized. The therapist will adjust the
amount of support for the mAAN. The rAAN algorithm automatically adjust the
amount of support (based on the error compared to a reference trajectory).
Walking quality and EMG signals will be measured during walking. In addition to
this, the subjects will be asked to fill in questionnaires about the subjective
rating of the support with the rAAN and mAAN algorithms.

Subjects will walk in the rehabilitation device LOPES II. Three different
support algorithms could be applied. The robotic and manual assist-as-needed
(rAAN and mAAN) only support the affected phases of gait cycle using a variable
joint stiffness. The amount of support will be minimized maintaining safe and
good qualitative walking. The rAAN algorithm automatically adjust the amount of
support based on the error compared to a reference trajectory. For the mAAN,
the therapist determines the amount of support which the robot provides.
The other approach is the position-controlled algorithm. This algorithm will
support the entire gait cycle. The amount of support will be minimized
maintaining safe and good qualitative walking.

Walking in LOPES II aims at improving walking. During the study, subjects will
walk in LOPES II with different support algorithms and for measuring joint
motions and walking parameters. In the first part, healthy subjects will be
asked to walk 6 trials, each with a duration of approximately 3 minutes. Rest
between the sessions is possible. Patients will walk 2 trails for maximal 5
minutes. Preparation will take about an hour. For the second and fourth part
(sub)acute SCI and stroke subjects are asked to walk 2 sessions for a maximum
of 5 minutes per session on their comfortable walking speed. Duration is
velocity dependent. Rest between the sessions is possible. Preparation for
measurements and walking will take about half an hour. The training period (2
periods of 3 weeks) is part of the inpatient rehabilitation program. Subjects
will receive training in LOPES II, instead of Body Weight Supported Treadmill
Training (BWSTT). Pre, between, post and follow up measurements will take
approximately 2 hours. Subjects are allowed to take rest during the training
and between the different measurements. A physiotherapist will accompany all
tests and training sessions.