healthmedicinet

Study design

The study was approved by the University of California, Irvine, Institutional Review
Board. Qualified chronic stroke subjects participated in 12 sessions of BCI-FES therapy
for foot drop. Each subject first underwent the following baseline assessments: fast
gait speed, dorsiflexion active range of motion (AROM), six-minute walk distance (6MWD),
and Fugl-Meyer leg motor (FM-LM) score. The BCI-FES therapy was then administered
at a rate of three one-hour sessions per week over the course of four weeks. The system
was designed to detect when the subject was trying to dorsiflex his/her paretic foot,
using EEG, and deliver electrical stimulation to the appropriate deep peroneal nerve.
Neurological and functional assessments were performed immediately prior to every
third session, as well as one week and four weeks after the 12
^th
session. The schedule of activities is summarized in Fig. 1. A before-and-after comparison was used to determine if any of the outcome measures
deteriorated significantly.

Fig. 1. Schedule of activities for each subject. Interval Assessments 1, 2, and 3 were performed
immediately before BCI-FES therapy sessions 4, 7, and 10, respectively

Subject recruitment

Advertisements were used to recruit stroke survivors (6 mo post-stroke) with gait
impairment that included foot drop. The entry criteria were: (1) age 18, (2) ability
to walk ?10 m without the use of an ankle-foot orthosis or assistance by another person
(walking aids, e.g. a cane or walker, were permitted but had to be used consistently),
(3) sufficient cognitive and language function to follow study-related instructions
reliably. The exclusion criteria were: (1) absence of a tibialis anterior response
to FES, (2) inability to tolerate FES, (3) presence of electronic implants, (4) severe
spasticity or plantarflexion contracture (Modified Ashworth scale?=?4). Although subjects
were allowed to continue their personal exercise programs, initiating new rehabilitative
treatments or physiotherapies during the study would result in exclusion from the
study.

BCI-FES dorsiflexion therapy

Qualified subjects underwent 12 one-hour long sessions of the experimental BCI-FES
dorsiflexion therapy, generally performed at a rate of three sessions per week across
four weeks. Before each session, the subjects were fitted with an EEG cap (10-10 International
Standard), and impedances were reduced to 10 k? for a fixed 32-channel set. To provide
the system with training EEG data, subjects followed 100 alternating six-second-long
cues to relax or dorsiflex the paretic foot. Subjects were visually monitored for
mirror movements (involuntary movements of a limb that are synchronous with the voluntary
movements of the contralateral limb 23]) and co-contractions at other limbs, and were asked to discontinue these if they
occurred. However, compliance with these instructions was not formally measured. The
training data were recorded at 256 Hz with a NeXus-32 bioamplifier (Mind Media, Herten,
Netherlands) to generate a session-specific decoding model for online operation (methods
in 24]–26]). This model could distinguish idling from dorsiflexion using EEG. Finally, surface
FES electrodes were placed over the proximal course of the deep peroneal nerve (paretic
side only), and stimulation parameters were adjusted to achieve ~15° dorsiflexion
from the neutral position without discomfort.

During each one-hour-long session, subjects performed as many online BCI-FES runs
as possible. Each run consisted of 10 alternating, contiguous 10-s-long dorsiflex/relax
cues, during which the BCI-FES system detected the subjects’ intention to dorsiflex
(or relax) from EEG and correspondingly provided (or withheld) stimulation. The accuracy
of BCI operation was calculated as the percentage of correctly determined BCI-states
at a rate of four decisions per second (decisions were calculated every 0.25 s based
on the most recent 0.75 s of EEG data 24]).

Subjects were informed of the possibility of erroneous recognition of the idle and
dorsiflexion states by the BCI. If FES was erroneously delivered when no movement
was intended, subjects were instructed to ignore the stimulation and continue to relax;
when movement was intended but no FES was delivered, subjects were instructed to continue
to attempt dorsiflexion. Mirror movements and co-contractions were monitored visually.
Brief breaks were provided between runs, or when requested, to prevent fatigue of
the deep peroneal nerve and tibialis anterior muscle.

Neurological and functional assessments

Neurological and functional measurements (fast gait speed, dorsiflexion AROM, 6MWD,
and FM-LM score) were performed before, during, and after the BCI-FES therapy, and
are defined below:

Gait Speed: Fast gait speed 27] was measured 28] for the middle 6-m section of a 10-m walkway. This test was repeated 5 times at each
assessment, and the average speed was calculated 29]. AFOs were removed, but walking aids such as walkers and canes were allowed. If a
subject did use a walking aid at their first baseline assessment, they were asked
to continue to use the same device throughout the rest of the study.

Dorsiflexion AROM: The subject was placed in a seated position with the knee flexed at 90° and the tibial
shank perpendicular to the ground. A goniometer was used to measure the AROM at the
ankle as the subject dorsiflexed, using standard technique 30].

6 min Walk Distance: Assessed as the distance that subjects can ambulate (at a safe, casual speed) in
6 min 31]. AFOs were removed, but walking aids were permitted.

Fugl-Meyer Leg Motor Score: Assessed using the FM measurement system as defined in 32], 33].

Although chronic stroke subjects are assumed to have reached a plateau in terms of
spontaneous behavioral recovery 34], 35], three baseline assessments (see Fig. 1) were performed to account for day-to-day variance 36]. Assessments were also performed immediately prior to every third BCI-FES session
and again one week and four weeks after the 12
^th
session. Subjects were also asked to maintain a fall diary throughout the study. The
results of this diary were documented at each weekly functional assessment.

Outcome measures

All pre-stated outcome measures focused on safety. The primary outcome was the proportion
of subjects who demonstrated a deterioration in gait speed ?0.16 m/s at either the
one-week or four-weeks post-therapy assessment. This threshold was chosen as it may
be associated with a change in the modified Rankin Scale (mRS) for the post-stroke
population 37]. The secondary outcome measures included the proportion of subjects who experienced
a significant deterioration in dorsiflexion AROM, 6MWD, and FM-LM score at either
the one-week or four-weeks post-therapy assessment. Deterioration of dorsiflexion
AROM is defined as a ?2.5° decrease from average baseline. There is no established
minimum clinically important change in dorsiflexion AROM, so this threshold was chosen
as it represents the minimal detectable difference 38]. The minimum clinically significant change in 6MWD is 20 % 39], while that of the FM-LM score is hypothesized to be 10 % 40].

Post-Hoc analyses

Additional analyses were performed to determine the proportion of subjects who demonstrated
a detectable increase in gait speed (?0.06 m/s 37], 39]), dorsiflexion AROM (?2.5°), 6MWD (?10 % 41]), and FM-LM score (?10 %) from average baseline at both post-therapy assessments.
Detectable changes are not necessarily clinically important, so the proportion of
subjects that experienced a clinically significant increase in gait speed (?0.16 m/s)
and 6MWD (?20 %) was also calculated. Furthermore, the effect of BCI-FES therapy on
gait speed, dorsiflexion AROM, 6MWD, and FM-LM was determined using independent linear
mixed models (LMMs) with outcomes (gait speed, etc.) as a function of therapy group
(pre-therapy baseline or post-therapy) with the random effects being subjects (and
their interaction with therapy group) and repetitions within subjects (and their interaction
with therapy group).

The training EEG datasets were analyzed for each session to determine if subjects
experienced any brain changes throughout the course of the therapy. To reveal any
spatial changes associated with therapy, the dorsiflexion-related importance (the
?-measure defined in 42]) of each EEG channel was plotted for all training sessions (details in Additional
file 1). Additionally, changes in event-related synchronization (ERS) and event-related
desynchronization (ERD) throughout the therapy were analyzed as follows. The most
important channel (highest average ? across all sessions) was identified for each
subject, and the EEG data from this channel were aggregated by week. The median ERS
and ERD at each frequency (8–30 Hz in 2 Hz bins) were calculated for weeks two through
four (Equations 2 and 3 in Additional file 1) and compared to their respective week-one value using repeated Mann–Whitney U tests
(Bonferroni corrected ??=?0.01).