Participants
The study was conducted at Coathill Hospital (NHS Lanarkshire), UK. Adults (over 18 years old) who had suffered a hemiplegic stroke from 1 week to 12 months and were medically stable and fit for gait rehabilitation were eligible to participate in the study. Patients who had severe cognitive impairment; significant hip, knee, or ankle contracture; walking difficulty before stroke; or contraindication for FES (e.g. using cardiac pacemaker or having metal implant under electrode sites) were excluded. Participants were enrolled prospectively, and written consent to participate in the study was obtained. This paper reports the development of the system and its feasibility with 6 participants who finished the training programme. The project is ongoing with a targeted sample size of 20 cases. The study was approved by NHS West of Scotland Research Ethics Committee (Reference: 17/WS/0245) and was registered in ClinicalTrials.gov (Reference: NCT03348215). R&D approval was granted by NHS Lanarkshire (R&D ID: L18006).
Training programme
All participants received up to 20 min of enhanced treadmill training session instead of their over-ground gait training session once or twice a week for 6 weeks. During training sessions, any ankle foot orthosis they were prescribed was removed. Participants could request pauses in the training session if needed. The treadmill was equipped with front and side bars from which support could be gained if required.
Research equipment and technology
Treadmill and body-weight supported system
The study used an N-Mill treadmill (Model: N-Mill 1 N75, ForceLink B.V., Culemborg, Netherlands) that can adjust its speed to match the user in other words operated in self-paced mode using D-Flow software. A harness and pneumatic body-weight supported system (PneuMex) was also provided (Fig. 2). The amount of body-weight support was adjusted to take into account the walking ability of each participant. However, all participants wore the harness attached to the suspension system to prevent falls.
3D-MoCap system
The 3D-MoCap consisted of a 6-camera motion analysis system (Bonita, Vicon Inc., Oxford, UK) which was installed around the treadmill. A cluster based method for gait analysis with pointer anatomical calibration was used [20]. The clusters of markers were placed on the pelvis, both thighs, both shanks, and both feet for the whole training session. Vicon Tracker software (Vicon Motion Systems, Oxford, UK) was used to identify cluster sets as objects, and to stream marker trajectory information into D-Flow (Motekforce Link, Netherlands). A 3D visualisation package D-flow, then, manipulated the input data from the 3D-MoCap using scripts written in the Lua programming language or using the provided D-Flow modules. The D-flow application presented the movement data as an avatar to the subjects on a large TV screen at the head of the treadmill. The anatomical calibration, the calculation of joint kinematics/gait parameters, and the phases of the gait cycle were carried out by D-flow at 60 Hz in real time.
The study used the toe marker trajectory on both sides to establish the phases of the gait cycle on the hemiplegic side. For treadmill walking, the stance phase was determined when the toe marker moved backward whereas the swing phase was determined when the marker moved forward. In addition, the stance phase was divided into 3 subphases: first double limb support (1DS); single limb support (SS); second double limb support (2DS) while the swing phase was divided in to 2 subphases: early swing (ESW) and late swing (LSW) as shown in Fig. 3.
FES
Dual-channel surface electrical stimulators (NeuroTrac® Rehab, Model number: ECS305A) were used in the present study. The electrical impulse was an asymmetrical, rectangular bi-phasic waveform. The commercial stimulator had a pushbutton remote switch to control the stimulation. When the remote switch was connected to the device, the electrical stimulation could be stopped and re-started by pressing the pushbutton switch. In the study, the pushbutton remote switch was removed and replaced with a computerised switch under software control. The stimulators were not modified, and were used in the custom mode (P15 mode) with 300 μs of pulse duration, 30 Hz of pulse frequency, and 0.1 s of ramp up/down. Amplitude (mA) was maximised to reach appropriate muscle function without pain or discomfort. Self-adhesive reusable skin electrodes size 50 × 50 mm were used individually for each subject.
For the FES software switch control system, an Arduino board was used. It had digital pins that could be outputs used to control the stimulators. The remote-control socket of the electrical stimulators was connected to the output pins of the Arduino producing a series of computer-controlled ON/OFF switches. The HIGH and LOW output are then similar to pressing and un-pressing the original remote switch. The Arduino board received real-time information (a number from 1, 2, 3, 4 or 5) from the computer giving the phase of the gait cycle (1 = 1DS; 2 = SS; 3 = 2DS; 4 = ESW; 5 = LSW) as calculated by D-flow. These output values were used to trigger the FES devices at the targeted time.
Up to 4 FES devices could be used for: Pre-tibial muscle (FES1) for reducing drop foot problem during SW; Gastro-soleus muscle (FES2) for restraining the tibia’s progression during SS and 2DS; Quadriceps (Vastus Medialis/Lateralis) muscle (FES3) and Hamstring (Semitendinosus/Biceps femoris [long head]) muscle (FES4) for improving hip and knee stability during 1DS (Fig. 4). The treating physiotherapist judged which FES devices were used to support gait retraining.
Visualisation feedback
Within the D-flow software, the animation was constructed and displayed in the “DSR” visualisation window. The dynamic visualisation of lower-limb kinematics was developed by linking segments with common joint centres (Fig. 5) [20]. Cylindrical objects were used to create each segment, and spherical objects were placed at each joint centre. A separate head, trunk, and pelvis was included in the Avatar. This visualisation was used for real-time visual feedback. Moreover, step length of both sides and step ratio were calculated and shown on the screen in real time. Step length was the distance in the antero-posterior direction between both toe markers at initial foot contact and step ratio was hemiplegic step length divided by the non-hemiplegic one. The distance covered and time spent walking were also shown on the screen.
Outcome measures
Number of training sessions attended, and training duration were used to assess the feasibility of enhanced treadmill training. In addition, joint kinematics in the sagittal plane of walking with and without FES support were evaluated to determine whether the developed FES system worked well in clinical practice. Moreover, walking speed from a 10-m walk test and functional mobility assessed by Rivermead Mobility Index (RMI) [21,22,23] were collected before and after treatment. Patients’ feedback based on a structured questionnaire was collected after patients completed their training programme.
Statistical analysis
Descriptive statistics were obtained including mean with standard deviation or median with interquartile range for continuous data (e.g. age, time since stroke onset), and number and percentage for categorical data (e.g. gender, hemiplegic side). The statistical difference of continuous outcomes (e.g. gait speed, joint kinematics) between 2 sets of dependent data (e.g. walking with FES and without FES) was examined using Wilcoxon signed-rank test given the small group size.