Abstract

Functional electrical stimulation (FES) is often used in poststroke gait rehabilitation to address decreased walking speed, foot drop, and decreased forward propulsion. However, not all individuals experience clinically meaningful improvements in gait function with stimulation. Previous research has developed adaptive functional electrical stimulation (AFES) systems that adjust stimulation timing and amplitude at every stride to deliver optimal stimulation. The purpose of this work was to determine the effects of a novel AFES system on functional gait outcomes and compare them to the effects of the existing FES system. Twenty-four individuals with chronic poststroke hemiparesis completed 64-min walking trials on an adaptive and fixed-speed treadmill with no stimulation, stimulation from the existing FES system, and stimulation from the AFES system. There was no significant effect of stimulation condition on walking speed, peak dorsiflexion angle, or peak propulsive force. Walking speed was significantly faster and peak propulsive force was significantly larger on the adaptive treadmill (ATM) than the fixed-speed treadmill (both p < 0.0001). Dorsiflexor stimulation timing was similar between stimulation conditions, but plantarflexor stimulation timing was significantly improved with the AFES system compared to the FES system (p = 0.0059). Variability between and within subjects was substantial, and some subjects experienced clinically meaningful improvements in walking speed, peak dorsiflexion angle, and peak propulsive force. However, not all subjects experienced benefits, suggesting that further research to characterize which subjects exhibit the best instantaneous response to FES is needed to optimize poststroke gait rehabilitation using FES.

Introduction

Stroke is one of the most common causes of disability in the United States [1] and often has detrimental long-term effects on mobility. Hemiparesis, where one side of the body loses motor control and function, often leads to decreased walking speed [24], foot drop [5], and decreased forward propulsion [3,6], which are correlated to increased risk of falls and mortality [7]. Rehabilitation programs often seek to restore walking speed through physical therapy, strength training, and gait training [2,8,9], but may not be successful [2,9].

Functional electrical stimulation (FES) is a therapeutic intervention where electrical impulses are applied to muscles to augment reduced motor control and restore function [2,5,9]. Stimulation-driven increases in dorsiflexor and plantarflexor muscle contraction often result in faster walking speed [2]. During dynamic tasks like walking, the timing of FES delivery is controlled by a series of footswitches placed on the soles of the shoes to determine the phase of gait. FES to the dorsiflexor (DF) muscles can reduce foot drop [5], while FES to the plantarflexor (PF) muscles can increase propulsion [10], thus improving typical gait impairments in individuals poststroke [3,11,12]. Particularly when combined with fast walking, FES can result in meaningful improvements in walking speed and propulsion [1315], but FES is not always effective for all subjects [13,16] and changes may not be retained after training [14,17].

In an attempt to make FES an effective rehabilitation tool for all individuals poststroke, several research groups have developed adaptive FES (AFES) protocols that iteratively adjust stimulation parameters during a walking trial [1825]. While these studies found improved peak dorsiflexion [18,2123] and inversion/eversion [19,21] during swing phase as well as increased walking speed [20], most of these groups did not include PF stimulation. The few adaptive stimulation systems that do include stimulation to the PF muscles are either based on electromyography signals [20] that may not be directly correlated to walking speed [6,2628], or validated their stimulation system on healthy young adults and did not test in a poststroke population [29]. Additionally, AFES systems have not been evaluated on an adaptive or variable speed treadmill, which may allow subjects to immediately leverage stimulation-driven increases in propulsion to increase walking speed.

To address limitations with previous AFES systems and determine the effects of variable stimulation amplitudes on an adaptive treadmill (ATM), we developed a novel AFES system that iteratively updates stimulation amplitudes to the DF and PF muscles at each stride based on measured dorsiflexion angle and peak propulsive force [30]. Additionally, the footswitch conditions for PF stimulation delivery were updated to decrease the frequency of PF stimulation delivery at the incorrect time. The purpose of this study is to evaluate the effect of the novel AFES system on functional gait parameters in individuals poststroke. Because we believe that the AFES system would generate stronger muscle contractions than the existing FES system, we hypothesized that:

  1. Walking speed will increase when walking on the adaptive treadmill with AFES compared to the existing FES system or no stimulation.

  2. Peak dorsiflexion angle during swing phase will increase with AFES compared to the existing FES system or no stimulation.

  3. Paretic peak propulsive force will increase with AFES compared to the existing FES system or no stimulation.

We also believed that the updated footswitch conditions in the AFES system would result in improved stimulation timing. Specifically, we hypothesized that stimulation will be delivered with greater accuracy relative to the gait cycle with the AFES system.

Materials and Methods

Subjects.

Twenty-four individuals with chronic poststroke hemiparesis (Table 1) were recruited through the University of Delaware’s Stroke Registry database. Subjects were required to be able to walk for 6 min and have passive ankle dorsiflexion to at least neutral and hip extension to at least 5 deg, in addition to having no falls in the previous 6 months and not currently participating in physical therapy for the lower body. Other inclusion and exclusion criteria were the same as in Ray et al. [31]. The study was approved by the University of Delaware’s Institutional Review Board and participants provided informed consent.

Table 1

Participant demographics, presented as mean±standard deviation

Gender9 M | 15 F
Age (years)63±10
Paretic side15 L | 9 R
Months since stroke102±72
Height (m)1.70±0.08
Weight (kg)81.21±18.09
Overground speed (m/s)1.02±0.30
6 MWT distance (ft)1165.41±338.72
Gender9 M | 15 F
Age (years)63±10
Paretic side15 L | 9 R
Months since stroke102±72
Height (m)1.70±0.08
Weight (kg)81.21±18.09
Overground speed (m/s)1.02±0.30
6 MWT distance (ft)1165.41±338.72

Adaptive Functional Electrical Stimulation System.

A Digitimer DS8R stimulator is connected to a D188 Remote Electrode Selector (Digitimer Ltd., Welwyn Garden City, UK), which allows stimulation to be delivered through two sets of electrodes, one for the DF and one for the PF muscles. A data acquisition board (USB-6341, National Instruments, Austin, TX; 600 Hz) controls stimulation amplitude, triggering stimulation on and off, and which set of electrodes is selected. The data acquisition board reads analog data from a set of bilateral footswitches (Noraxon, Scottsdale, AZ) on the heel and toe of each foot, which are used to determine when stimulation should be delivered. A laptop running the stimulation code in MATLAB (MathWorks, Natick, MA) is connected to an 8-camera motion capture system (Qualisys AB, Göteborg, Sweden; 100 Hz) and receives kinematic data from the marker set and six degree-of-freedom forces and moments from the instrumented treadmill (Bertec Corp., Columbus, OH; 2000 Hz) to calculate real-time gait biomechanics.

Experimental Protocol.

This study was performed over two separate visits at least 24 h apart to mitigate any effects of fatigue and no more than 4 weeks apart. All subjects completed both sessions. During the first visit, subjects provided informed consent and were evaluated for eligibility in the study using the 6-min walk test and passive range of motion measures. Dorsiflexion was assessed while seated and hip extension was assessed during static standing. Then, subjects were familiarized with the experimental protocol and lab equipment, including the ATM [32] and the AFES system [30]. Subjects were given up to 5 min of familiarization with the ATM to ensure they were comfortable and able to maintain a consistent speed. Sixteen reflective motion capture markers were placed on each leg, 8 on anatomical landmarks and 8 on rigid shells. Then, surface FES electrodes (Axelgaard Manufacturing Co., Fallbrook, CA) were placed over the anterior tibialis (2 in. by 2 in.) and medial gastrocnemius (3 in. by 5 in.) of the paretic leg. Electrode placement was determined based on individual subject muscle geometry and adjusted based on the motor response under the guidance of a licensed physical therapist. Minimum and maximum stimulation amplitudes were set at the subject’s sensory threshold and 5 mA below the pain threshold, respectively. The initial FES amplitude was set based on the desired motor response for each muscle. The motor response for DF stimulation was dorsiflexion to a neutral ankle angle while the subject was seated, and the motor response for the PF stimulation was set while the subject was standing with most of the body weight on the nonparetic limb until the knee began to flex and the heel began to lift off the ground [10,16,18]. Footswitches were placed bilaterally on the heel and toe to control the timing of the stimulation [10,33]. The stimulation was composed of a monophasic variable frequency train of 3 pulses at 200 Hz followed by 10 pulses at 30 Hz, as this pattern has been found to reduce fatigue while maintaining motor response [10,33]. The pulse width was 200 μs. Each train of stimulation lasts approximately 300 ms, and the variable frequency train repeats as long as the footswitch conditions are met.

During the walking trials, DF stimulation was delivered during paretic swing phase and PF stimulation was delivered during paretic terminal stance [10,16,34]. In both the FES and AFES conditions, paretic swing phase was detected when any nonparetic footswitches were in contact with the ground and when no paretic footswitches were in contact with the ground. In the FES condition, paretic terminal stance was detected when any nonparetic footswitches and the paretic toe footswitch were in contact with the ground. In the AFES condition, paretic terminal stance was detected using sequential footswitch conditions to ensure that the AFES system delivers PF stimulation only during terminal stance and not during a forefoot strike, which is common with foot drop [30]. The paretic heel or foot conditions must precede onset of the paretic toe footswitch. For both FES and AFES, the footswitch conditions for stimulation delivery did not change over time. However, the AFES system incorporated updated footswitch conditions that we hypothesized would better detect gait events in a variable poststroke population.

In the FES condition, stimulation amplitudes were constant at the initial amplitude that was set based on the motor response. In the AFES condition, stimulation amplitudes were updated iteratively at each stride based on measured dorsiflexion angle and peak propulsive force asymmetry [30]. The measured dorsiflexion angle was calculated as the dorsiflexion angle at initial contact to capture foot drop [3538]. The peak propulsive asymmetry was calculated as the nonparetic peak propulsive force minus the paretic peak propulsive force divided by the sum. Less dorsiflexion than the goal dorsiflexion angle and positive peak propulsive force asymmetry (nonparetic force greater than paretic force) resulted in increased stimulation amplitudes. Vice versa, more dorsiflexion than the goal dorsiflexion angle and negative peak propulsive force asymmetry (paretic force greater than nonparetic force) resulted in decreased stimulation amplitudes. The goal dorsiflexion angle was set to a neutral angle for each subject during quiet standing. The propulsion goal was set to zero peak propulsive force asymmetry. Subjects were given up to 5 min of walking with stimulation for familiarization.

During the second visit, following setup identical to the first session, participants walked with and without stimulation on the fixed-speed (FSTM) and adaptive treadmills. Trials without stimulation were performed first to remove the possibility of stimulation-induced fatigue [39], and then trials with stimulation were performed in a randomized order. All trials were 4 min long, and trials with stimulation had either FES or AFES alternating on and off in 1-min intervals for the entirety of the 4-min trial [26]. At the beginning of the trial, subjects took up to 1 min to reach their comfortable walking speed. On the adaptive treadmill, the speed changed based on user step length, propulsive force, and position [32], and subjects verbally indicated when they had achieved their comfortable speed and confirmed that they could maintain that speed for 4 min. On the fixed-speed treadmill, the treadmill was started at the subjects’ overground walking speed and adjusted in 0.05 m/s increments until the subjects verbally indicated that they had achieved their comfortable speed. Once subjects reached their comfortable speed, the stimulation was started.

Data Analysis.

Kinetic and kinematic data were filtered with a fourth-order low-pass Butterworth filter with a cutoff frequency of 30 Hz and 6 Hz, respectively, in Visual 3D (C-Motion Inc., Germantown, MD). Paretic peak dorsiflexion angle was calculated as the minimum dorsiflexion angle during swing phase. Paretic peak anterior ground reaction force (AGRF) was calculated as the maximum anterior-posterior ground reaction force during stance phase. System performance was characterized by the percentage of strides in each trial that had stimulation delivered at the correct time (swing phase for DF stimulation and terminal stance for PF stimulation), no stimulation delivered at all, stimulation delivered at both the correct and incorrect time, and stimulation delivered at only the incorrect time (terminal stance for DF stimulation and early stance for PF stimulation) [30].

All biomechanical outcome measures were assessed for normality using the Shapiro–Wilk test and compared between conditions using a two-way repeated measures ANOVA or Friedman’s test with a significance level of 0.05. Tukey posthoc testing was performed as needed. Stimulation timing data were assessed for normality using the Shapiro–Wilk test and compared between FES and AFES systems using Wilcoxon Signed Rank tests. Common language effect sizes were calculated for all comparisons between stimulation conditions [40].

Results

One subject was excluded from analysis because of an excessive number of crossover steps resulting in very few usable strides. The following analysis includes the remaining 23 subjects.

Walking speed was not significantly different between stimulation conditions (p = 0.86) but was significantly faster on the ATM than the FSTM (p < 0.0001, Table 2, Fig. 1). There was no interaction between treadmill and stimulation conditions (p = 0.95). Group average walking speed ranged from 0.70 m/s with no stimulation on the FSTM to 0.82 m/s with FES on the ATM, but all standard deviations were large (0.22–0.29 m/s). Individual walking speeds ranged from 0.21 m/s (subject 17, FES-ATM) to 1.4 m/s (subject 26, FES-FSTM).

Fig. 1
Walking speed with no stimulation (dark grey), FES (medium grey) and AFES (light grey) on the FSTM (left group) and ATM (right group). Error bars are±one standard deviation and individual subject data points are overlaid with connecting lines. There was no effect of stimulation condition on walking speed on either treadmill condition. Walking speed was significantly faster on the ATM than the FSTM (p < 0.0001).
Fig. 1
Walking speed with no stimulation (dark grey), FES (medium grey) and AFES (light grey) on the FSTM (left group) and ATM (right group). Error bars are±one standard deviation and individual subject data points are overlaid with connecting lines. There was no effect of stimulation condition on walking speed on either treadmill condition. Walking speed was significantly faster on the ATM than the FSTM (p < 0.0001).
Close modal
Table 2

Mean±standard deviation walking speed, peak dorsiflexion angle, and paretic peak propulsive force (AGRF) in all six experimental conditions

Stimulation condition
NoneFESAFES
Walking speed (m/s)FSTM0.7 ± 0.230.71 ± 0.240.71 ± 0.22Treadmill main effect: <0.0001
ATM0.81 ± 0.240.82 ± 0.290.81 ± 0.29
Stimulation main effect: 0.86
Dorsiflexion angle (°)FSTM6.3 ± 5.76.9 ± 5.76.8 ± 5.4Treadmill main effect: 0.82
ATM6.9 ± 5.57.1 ± 5.86.3 ± 5.9
Stimulation main effect: 0.42
Paretic peak AGRF (% BW)FSTM9.13 ± 5.149.15 ± 5.119.11 ± 4.96Treadmill main effect: <0.0001
ATM10.28 ± 5.8110.53 ± 6.5510.44 ± 6.14
Stimulation main effect: 0.88
Stimulation condition
NoneFESAFES
Walking speed (m/s)FSTM0.7 ± 0.230.71 ± 0.240.71 ± 0.22Treadmill main effect: <0.0001
ATM0.81 ± 0.240.82 ± 0.290.81 ± 0.29
Stimulation main effect: 0.86
Dorsiflexion angle (°)FSTM6.3 ± 5.76.9 ± 5.76.8 ± 5.4Treadmill main effect: 0.82
ATM6.9 ± 5.57.1 ± 5.86.3 ± 5.9
Stimulation main effect: 0.42
Paretic peak AGRF (% BW)FSTM9.13 ± 5.149.15 ± 5.119.11 ± 4.96Treadmill main effect: <0.0001
ATM10.28 ± 5.8110.53 ± 6.5510.44 ± 6.14
Stimulation main effect: 0.88

Statistically significant comparisons are bolded. Walking speed and peak AGRF differed between FSTM and ATM, with the ATM speed being significantly faster (p < 0.0001) and the ATM AGRF being significantly larger (p < 0.0001). There was no effect of stimulation condition on any outcome measure.

Peak dorsiflexion angle was not statistically significantly different between stimulation conditions (p = 0.42) or treadmill conditions (p = 0.82, Table 2, Fig. 2). There was no interaction between treadmill and stimulation conditions (p = 0.33). Average peak dorsiflexion angle was smallest with no stimulation on the FSTM and with AFES on the ATM (6.33 deg) and largest with FES on the FSTM (7.08 deg). Standard deviations were again large, greater than 5 deg in all conditions. The minimum peak dorsiflexion angle was −6.6 deg (subject 27, AFES-ATM), while the maximum peak dorsiflexion angle was 18.3 deg (subject 3, AFES-FSTM).

Fig. 2
Peak dorsiflexion angle during swing phase with no stimulation (dark grey), FES (medium grey) and AFES (light grey) on the FSTM (left group) and ATM (right group). Positive angles are dorsiflexion and negative angles are plantarflexion. Error bars are±one standard deviation and individual subject data points are overlaid with connecting lines. There was no significant effect of stimulation condition or treadmill condition.
Fig. 2
Peak dorsiflexion angle during swing phase with no stimulation (dark grey), FES (medium grey) and AFES (light grey) on the FSTM (left group) and ATM (right group). Positive angles are dorsiflexion and negative angles are plantarflexion. Error bars are±one standard deviation and individual subject data points are overlaid with connecting lines. There was no significant effect of stimulation condition or treadmill condition.
Close modal

Paretic peak propulsive force was not significantly different between stimulation conditions (p = 0.88), but was significantly different between treadmill conditions (p < 0.0001, Table 2, Fig. 3). There was no interaction between treadmill and stimulation conditions (p = 0.89). Paretic peak propulsive force was statistically significantly larger on the ATM (10.42% bodyweight (BW)) than the FSTM (9.13% BW, p < 0.0001). Peak propulsive force was smallest with AFES on the FSTM (9.11% BW) and largest with FES on the ATM (10.53% BW). There was a large amount of between-subject variation, reflected by the large standard deviations (4.96–6.55% BW) and the wide range between the smallest peak propulsive force (0.1% BW, subject 17, FES-ATM) and the largest peak propulsive force (23% BW, subject 19, FES-ATM).

Fig. 3
Peak propulsive force with no stimulation (dark grey), FES (medium grey) and AFES (light grey) on the FSTM (left group) and ATM (right group). Error bars are±one standard deviation and individual subject data points are overlaid with connecting lines. There was no significant effect of stimulation condition, but peak propulsive force was significantly larger on the ATM (p < 0.0001).
Fig. 3
Peak propulsive force with no stimulation (dark grey), FES (medium grey) and AFES (light grey) on the FSTM (left group) and ATM (right group). Error bars are±one standard deviation and individual subject data points are overlaid with connecting lines. There was no significant effect of stimulation condition, but peak propulsive force was significantly larger on the ATM (p < 0.0001).
Close modal

Stimulation timing to the DF muscles was slightly improved with the AFES system compared to the FES system, although not statistically significantly (Table 3, Fig. 4). Both the FES and AFES systems correctly delivered DF stimulation during swing phase to greater than 96% of strides, with the AFES system demonstrating a slight improvement. No strides were missing stimulation to the dorsiflexor muscles, and no strides had stimulation only applied at the incorrect time but not at the correct time. There was a small decrease in multiple instances of DF stimulation deliveries during terminal stance with the AFES system (2.29±6.73%) compared to the FES system (3.70±10.76%, p = 0.052). Both standard deviations are somewhat large due to the variance between subjects ranging from 0% of strides with multiple instances of stimulation to 63% of strides with multiple instances of DF stimulation (subject 2, FES-ATM). For the plantarflexor muscles, stimulation timing was also improved with the AFES system compared to the FES system (Table 3, Fig. 4). The FES system delivered PF stimulation at the correct time to 78.93±34.51% of strides, while the AFES system delivered correctly timed PF stimulation to 89.48±22.57% of strides, indicating a substantial but not statistically significant improvement (p = 0.19). The FES system did not deliver PF stimulation at all for 18.84±33.87% of strides, while the AFES system significantly improved this percentage and only missed delivering stimulation to 6.22±21.09% of strides (p = 0.0059). The FES system delivered multiple instances of PF stimulation to 2.17±10.31% of strides, while the AFES system had slightly worse performance with 4.30±10.19% of strides with incorrectly timed PF stimulation, although this comparison was not statistically significant (p = 0.10). Finally, in both the FES (0.06±0.00%) and AFES (0.00±0.00) systems, very few strides received stimulation at only the incorrect time.

Fig. 4
FES versus AFES system performance for DF (top) and PF (bottom). Bars indicate mean±standard deviation. Both stimulation systems delivered DF stimulation at the correct time to greater than 96% of strides. The AFES system had slightly fewer incorrect DF stimulation deliveries and both systems had zero strides missing DF stimulation or with DF stimulation delivered at only the incorrect time. The AFES system delivered PF stimulation at the correct time to substantially more strides than the FES system, although not significantly (89.48% versus 78.93%, p = 0.19). The AFES system delivered PF stimulation at the incorrect time slightly more than the FES system, although this comparison was not statistically significant.
Fig. 4
FES versus AFES system performance for DF (top) and PF (bottom). Bars indicate mean±standard deviation. Both stimulation systems delivered DF stimulation at the correct time to greater than 96% of strides. The AFES system had slightly fewer incorrect DF stimulation deliveries and both systems had zero strides missing DF stimulation or with DF stimulation delivered at only the incorrect time. The AFES system delivered PF stimulation at the correct time to substantially more strides than the FES system, although not significantly (89.48% versus 78.93%, p = 0.19). The AFES system delivered PF stimulation at the incorrect time slightly more than the FES system, although this comparison was not statistically significant.
Close modal
Table 3

FES versus AFES system performance

Stimulation condition
FESAFESp-value
DF stimulation% Correct96.30 ± 10.7697.71 ± 6.730.052
% Missing0.00 ± 0.000.00 ± 0.001
% Multiple3.70 ± 10.762.29 ± 6.730.052
% Incorrect0.00 ± 0.000.00 ± 0.001
PF stimulation% Correct78.93 ± 34.5189.48 ± 22.570.19
% Missing18.84 ± 33.876.22 ± 21.090.0059
% Multiple2.17 ± 10.314.30 ± 10.190.10
% Incorrect0.06 ± 0.390.00 ± 0.001
Stimulation condition
FESAFESp-value
DF stimulation% Correct96.30 ± 10.7697.71 ± 6.730.052
% Missing0.00 ± 0.000.00 ± 0.001
% Multiple3.70 ± 10.762.29 ± 6.730.052
% Incorrect0.00 ± 0.000.00 ± 0.001
PF stimulation% Correct78.93 ± 34.5189.48 ± 22.570.19
% Missing18.84 ± 33.876.22 ± 21.090.0059
% Multiple2.17 ± 10.314.30 ± 10.190.10
% Incorrect0.06 ± 0.390.00 ± 0.001

Mean ± standard deviation percentage of in each category of stimulation timing. Statistically significant comparisons are bolded. The AFES system slightly increased correct DF stimulation deliveries and decreased strides with multiple instances of stimulation, although not significantly. The AFES system increased the number of strides with correctly timed PF stimulation, but not statistically significantly), although the percentage of strides missing stimulation was significantly decreased with AFES (p = 0.0059). The AFES system had slightly more strides with multiple instances of PF stimulation, but not statistically significantly.

Discussion

The purpose of this study was to evaluate the instantaneous effect of the novel AFES system on functional gait parameters in individuals poststroke and compare the effects to the existing FES system. Overall, there were no statistically significant effects of the stimulation, either FES or AFES, on walking speed, peak dorsiflexion angle, or paretic peak propulsive force. However, there was substantial intersubject variability where some subjects responded to the stimulation, either FES or AFES, and some subjects did not. Subjects were classified as responders for each of the three main outcome measures if they increased that metric with FES or AFES relative to the no stimulation condition. Subjects were classified as nonresponders if the value in the FES or AFES conditions did not exceed the value in the no stimulation condition.

Walking Speed.

It was hypothesized that walking speed would increase on the ATM with AFES compared to no stimulation. This hypothesis was not supported by the data, although walking speed was significantly faster on the ATM than the FSTM by approximately 0.1 m/s (Table 2, Fig. 1). This aligns with previous research in healthy young adults and stroke survivors where speed increases on the ATM relative to the FSTM [31,32]. Because of the ATM’s ability to promote faster walking speed and natural gait variability [41,42], walking on the ATM may be beneficial for poststroke gait rehabilitation if used in a long-term therapeutic intervention. Additionally, combining the ATM with stimulation from either the FES or AFES system may provide additional benefits.

Walking speed did not differ between any stimulation conditions, which contrasts the hypothesized result and previous research [2,34,43]. The effect size for the comparison between no stimulation on the ATM and FES on the ATM was 56.52%, indicating a small effect where some people walked faster with stimulation than without. This conclusion is further supported by the moderate effect size comparing no stimulation on the ATM and AFES on the ATM (65.22%), suggesting that more people increased their walking speed with AFES than FES. However, the effect of stimulation on walking speed was inconsistent. The average overground walking speed across all participants was 1.01 m/s, which is considerably faster than many previous studies in individuals poststroke [4,4446]. It is possible that most individuals included in this study were higher functioning than subjects in previous studies who increased walking speed when walking with stimulation. For example, the single subject in Chen et al. increased walking speed from 0.22 to 0.46 m/s, which is much slower than the average walking speed in this study [20]. Consequently, the subjects in this study may not have had as much room for improvement, resulting in a ceiling effect. While many people may instantaneously benefit from stimulation with either the FES system or the AFES system, the larger instantaneous effect size with the AFES system suggests that there may be more potential for improvement following a long-term intervention with AFES compared to FES.

To further examine the effects of the stimulation, subjects were categorized into groups based on which stimulation condition resulted in the fastest walking speed. Only ATM trials were included in this analysis because of the ability to instantaneously increase or decrease walking speed. Nine out of 23 subjects (39%) had the fastest walking speed with no stimulation (0.79±0.22 m/s, Table 1 available in the Supplemental Materials on the ASME Digital Collection). This average speed is relatively fast for individuals poststroke and would result in these participants being classified as “moderate,” “fast,” or “least-limited community ambulators” using a classification mechanism similar to that of other studies [4547]. Therefore, it is possible that these individuals did not walk faster with stimulation because of a ceiling effect where they were already walking as fast as was comfortable. However, one subject in this group (subject 17) walked at 0.56 m/s overground and 0.54 m/s with no stimulation, but slowed considerably to 0.21 m/s with FES and 0.36 m/s with AFES. This subject may have been unfamiliar or uncomfortable with the stimulation and altered their gait patterns to avoid experiencing stimulation for extended periods of time. This conclusion is supported by the increase in stance time asymmetry from 3.80±1.69% with no stimulation to 5.75±2.39% with FES and 5.53±2.19% with AFES. This increase indicates that the subject off-loaded their paretic leg and compensated by spending more time on the nonparetic leg. Increased stance time asymmetry has been correlated with slower walking speed [48], which would explain why this subject walked more slowly with stimulation. Additionally, this temporal alteration may have reduced the subject’s ability to achieve a large trailing limb angle and generate sufficient propulsion because they did not spend enough time in paretic stance [3,44].

Ten of 23 subjects (43%) performed best in the FES condition, and increased their walking speed by 0.08±0.06 m/s (Table 1 available in the Supplemental Materials). The average walking speed in this group was 0.84±0.29 m/s, which is again fairly fast for individuals poststroke and may have contributed to a similar ceiling effect that explains the small increase. However, these participants were able to increase their walking speed to some extent, likely due to stimulation-driven increases in paretic peak propulsive force, which increased by 1.29±1.62% BW. This increase in peak propulsive force exceeds the minimal detectable change for within session changes in peak propulsive force (0.80% BW) [49] and suggests that FES to the PF muscles may result in clinically meaningful increases in propulsion that translate to increases in walking speed.

Only four of 23 subjects (17%) had the fastest walking speed in the AFES condition, with an average walking speed of 1.00±0.27 m/s (Table 1 available in the Supplemental Materials on the ASME Digital Collection). These subjects increased their speed by 0.22±0.10 m/s compared to the no stimulation condition where their average speed was 0.77±0.20 m/s. This increase in walking speed exceeds the minimal detectable change for walking speed, which is 0.15 m/s [46,5052], and was paralleled by increases in peak propulsive force of 1.85±1.55% BW, also exceeding the minimal detectable change [49]. Three of these four subjects walked progressively faster with FES and AFES compared to no stimulation, suggesting that the FES condition offered some benefit and that the AFES condition provided additional potential for improvement.

Peak Dorsiflexion Angle.

It was hypothesized that paretic peak dorsiflexion angle would increase with AFES compared to FES or no stimulation, but peak dorsiflexion angle was not statistically different between stimulation conditions at the group level (Table 2, Fig. 2). Effect sizes were minimal for the comparison between FES and no stimulation (52.17%), AFES and no stimulation (52.17%) and AFES and FES (52.17%) on the ATM. These results are similar to previous studies, where peak dorsiflexion was limited when dorsi- and plantarflexor stimulation were combined [10]. DF stimulation alone typically increases peak dorsiflexion [10,39], but when PF stimulation is used in conjunction with DF stimulation, ankle plantarflexion increases at toe-off [10]. Therefore, to achieve the same dorsiflexion angle during swing phase with increased plantarflexion at toe-off, the ankle must travel through a larger range of motion in the same amount of time. Even with the assistance of DF stimulation, this may be difficult for individuals poststroke with impaired motor control. Additionally, PF stimulation often continues into early stance as the system responds to toe-off to switch between PF and DF stimulation. During this brief transition period, dorsiflexion may have been impeded by the PF stimulation, creating additional resistance to dorsiflexion in early swing. Some subjects saw substantial improvements in peak dorsiflexion angle, with several subjects improving dorsiflexion by 3–4 deg, which exceeds the minimum detectable change threshold of 1 deg [49]. Of the subjects who increased dorsiflexion with stimulation, the average increase was 2.6±1.1 deg with FES and 1.9±1.1 deg with AFES. Both instantaneous increases exceed the minimal detectable change of 1 deg for kinematic variables [49], suggesting that DF stimulation may be beneficial for some participants when used as a therapeutic intervention. To further analyze the potential subject-specific effects, subjects were divided into three groups based on which stimulation condition resulted in the most dorsiflexion.

Nine of 23 subjects (39%) achieved the largest peak dorsiflexion angle without stimulation (Table 2 available in the Supplemental Materials). Of these nine people, seven of them had dorsiflexion angles greater than 5 deg with no stimulation and all nine of them had dorsiflexion angles greater than neutral with no stimulation, suggesting that most of these subjects may not have needed DF stimulation. Therefore, it is reasonable that these subjects did not respond by increasing dorsiflexion angle, as they already had sufficient dorsiflexion without the assistance of DF stimulation. Previous research with a similar AFES system found that peak dorsiflexion increased significantly compared to no stimulation [22]. However, their dorsiflexion angles with no stimulation were approximately 0.6±3.0 deg, substantially smaller than the 6.9±5.5 deg presented in this study, so they may not have experienced a similar ceiling effect. Future research will develop screening criteria to better select the subjects who are most likely to respond to the stimulation to ensure that all subjects receive the most benefit.

Five of 23 subjects (22%) had the largest dorsiflexion angle in the FES condition, with an average increase of 2.7±1.1 deg compared to with no stimulation (Table 2 available in the Supplemental Materials). This result suggests that DF stimulation may improve dorsiflexion in some individuals poststroke. Four of the five subjects had substantial dorsiflexion without stimulation, so it is unclear what distinguishes these subjects from the group that performed best with no stimulation, although the heterogeneity of the poststroke population likely played a role [53]. These subjects likely performed better with FES than AFES because all five subjects had improved PF stimulation timing with AFES (98.10±2.61% correct with AFES versus 38.61±33.56% correct with FES). Improved PF stimulation timing resulted in increased plantarflexion at toe-off, which may have impeded the subjects’ ability to generate sufficient dorsiflexion. Future research will perform additional analyses using baseline characteristics and individual subject responses to characterize why some subjects respond while others do not. With this knowledge, researchers and clinicians can select the most beneficial rehabilitation protocol for each individual.

Finally, nine subjects (39%) achieved the most dorsiflexion with AFES, increasing 1.9±1.3 deg over the no stimulation condition and 1.4±1.1 deg compared to the FES condition (Table 2 available in the Supplemental Materials). This improvement is similar to the increases seen in Dong et al., where young healthy subjects increased dorsiflexion by approximately 2 deg with their AFES system. Dong et al. had considerably smaller standard deviations than in this study, likely because of the increased heterogeneity in the individuals poststroke that participated in this study [29]. Of these nine subjects, six subjects had larger average stimulation amplitudes in the AFES condition than the FES condition, which likely resulted in increased dorsiflexion. Four of these subjects had negative peak dorsiflexion angles with no stimulation or FES, indicating significant foot drop, and the AFES system increased dorsiflexion by nearly 2 deg for these subjects. Not all subjects were able to achieve positive dorsiflexion angles even with the AFES system, but the AFES system still resulted in reduced foot drop. Overall, 14 of 23 subjects improved their peak dorsiflexion with stimulation, either from the FES system or the AFES system. Those who did not respond by increasing dorsiflexion likely did not need the assistance of the stimulation, so future work will examine if a long-term training paradigm could create lasting improvements in peak dorsiflexion angle.

Paretic Peak Propulsive Force.

It was hypothesized that paretic peak propulsive force would increase with AFES compared to the existing FES system or no stimulation. Paretic peak propulsive force was not significantly different between stimulation conditions, although there was a significant increase from the FSTM to the ATM (∼1.29% BW, Table 2, Fig. 3). This aligns with previous research where propulsion increased slightly on the ATM, likely due to increases in trailing limb angle [31]. Additionally, the increase in propulsion parallels the increase in walking speed, since propulsion and walking speed are highly correlated [6]. Chen et al. and Dong et al. did not include propulsive measures, but Chen et al. did compare electromyography between paretic and nonparetic gastrocnemius muscles before and after 12 weeks of training [20]. Chen et al. found that the paretic limb performed significantly worse than the nonparetic leg before training, but that there were no significant differences in gastrocnemius muscle activity following training, suggesting that long-term training with an AFES system may be beneficial for overall function and forward propulsion.

Propulsive force was expected to increase with PF stimulation, as seen in previous studies [10], and 16 out of 23 subjects (70%) increased propulsion with stimulation (Table 3 available in the Supplemental Materials). Nine subjects (39%) had the largest peak propulsive force with FES and seven subjects (30%) had the largest peak propulsive force with AFES. With FES, subjects increased their peak propulsive force by 1.45±1.48% BW. With AFES, subjects increased their peak propulsion by 1.69±1.29% BW. In both stimulation conditions, the subjects who increased their propulsive force experienced substantial improvements, exceeding the minimal detectable change for peak propulsion [49]. These results suggest that stimulation may be beneficial for some participants, and that AFES may be more beneficial than FES for some people but also that FES may be more beneficial for others. Both FES and AFES show instantaneous improvements, suggesting that both stimulation systems may have the potential for improving poststroke gait following a long-term intervention. The subjects with the largest peak AGRF with AFES had larger average PF stimulation amplitudes than with FES (+17.49±14.28 mA), which may have helped these subjects increase their propulsion more with AFES than FES. The subjects with the largest peak propulsive force with FES did have larger PF stimulation amplitudes with AFES than FES (+8.51±16.06 mA), but this increase was much smaller and may not have been sufficient to generate increased propulsion. Additionally, three subjects in the FES group had decreased PF stimulation amplitudes with AFES, suggesting that increased average PF stimulation amplitude may be related to peak propulsive force.

There were seven people (30%) who had the largest peak propulsive force with no stimulation, five of whom also had the fastest walking speed with no stimulation (Table 3 available in the Supplemental Materials on the ASME Digital Collection). The goal of the PF stimulation was to promote ankle plantarflexion and increase propulsion as in previous studies [10,16,26] because ankle plantarflexion moment is strongly correlated to propulsion [54]. However, stimulation applied to the medial gastrocnemius may also cause knee flexion, as this muscle is biarticular. Knee extension moments in terminal stance contribute to forward propulsion [55], so increases in knee flexion due to gastrocnemius stimulation may cause buckling at the knee and be associated with decreased forward propulsion. When examining the knee flexion angles for subject 17, the nonresponder from the walking speed analysis, there are substantial increases in knee flexion at the time of peak propulsive force with FES (14.02±3.46 deg) and AFES (17.20±2.27 deg) compared to without stimulation (12.56±2.37 deg). Both of these increases exceed the minimal detectable change of 1 deg for joint angles [49], and the increase from FES to AFES is likely due to larger PF stimulation amplitude with AFES (89.62 mA on average) than FES (65 mA). Overall, it is possible that the poor response of paretic peak propulsive force to PF stimulation in these seven subjects is due to altered knee flexion or other kinematic changes that reduce the transfer of PF stimulation to propulsion. Determining which subjects will increase propulsion with PF stimulation presents a clear avenue for future work.

Stimulation Timing.

It was hypothesized that stimulation timing would improve with the AFES system compared to the existing FES system. This hypothesis was partially supported by the decrease in the percentage of strides missing PF stimulation, although DF stimulation timing was similar between conditions (Table 3, Fig. 4).

Both the FES and AFES systems delivered correctly timed stimulation to the DF muscles during swing phase for greater than 96% of strides, demonstrating that the swing phase detection algorithm is fairly robust. However, for both stimulation systems, 6–7% of strides had DF stimulation delivered at both the correct time and the incorrect time in terminal stance. Although this erroneous stimulation delivery occurred in a small percentage of strides on average, several subjects had DF stimulation frequently delivered in terminal stance. Across all four stimulation conditions (FES and AFES on the FSTM and ATM), only five subjects (subjects 2, 10, 12, 17, and 25) had multiple instances of DF stimulation during more than 5% of strides. These 5 subjects had multiple instances of DF stimulation to an average of 21.43% of strides, and two of them (subjects 2 and 17) had multiple instances of DF stimulation to greater than 5% of strides in all four stimulation conditions. This result suggests that there may be sensitivity issues with the footswitches that can cause erroneous stimulation delivery, regardless of the stimulation system being used. If the paretic footswitches are not sensitive enough, they may turn off while the leg is still in terminal stance, causing DF stimulation to be delivered too early. Additionally, there is typically some residual DF stimulation into early stance while the FES or AFES system responds to initial contact, but if the paretic footswitches do not turn on until terminal stance, DF stimulation delivery may be prolonged until late into the gait cycle. Future work will seek to address these timing issues to further reduce the percentage of strides with DF stimulation delivered in terminal stance. For example, a few previous research groups have developed predictive stimulation timing algorithms that do not rely on footswitches to detect phases of gait [24,56]. Implementing a similar predictive timing algorithm in the AFES system would make the system pro-active instead of reactive in detecting transitions between phases of gait, which may improve stimulation timing.

There was a substantial but not significant increase in the percentage of strides with correctly timed PF stimulation in terminal stance with the AFES system (89%) compared to the FES system (79%), indicating that the improvements to the PF stimulation timing algorithm were successful at improving PF stimulation timing. The FES system uses an independent activation condition where PF stimulation is only delivered if the paretic toe footswitch is on and the paretic heel footswitch is off. As a result, the FES system is not robust to variable gait patterns, slight variations in footswitch placement or shoe sole structure, or overly sensitive footswitches that might result in the paretic heel footswitch remaining on during terminal stance. The AFES system uses a sequential activation condition where PF stimulation is delivered following a sequence of footswitch conditions, which allows the system to be more robust to variability during data collection and therefore improves the accuracy of PF stimulation timing [30]. Additionally, the AFES system significantly improved the percentage of strides with no PF stimulation, decreasing from 19% with the FES system to 6% with the AFES system. This improvement indicates the sequential activation condition can also improve the detection of terminal stance by accounting for variable gait patterns. Although the percentage of correctly timed PF stimulation improved greatly with the AFES system, there were 2–4% of strides with multiple instances of PF stimulation with both stimulation systems. Similarly, a very small portion of strides had stimulation delivered at only the incorrect time (0.06% with FES and 0% with AFES). While this represents a very small proportion of the total number of strides, we suspect that this is due to similar footswitch sensitivity issues as the erroneous DF stimulation, resulting in footswitch conditions that do not align with the physiologic phases of gait. Future work will address the sensitivity issues or pursue other stimulation timing algorithms to further reduce erroneous stimulation delivery.

Previous research with similar AFES systems had considerably better gait event detection, with five phases of gait being detected with 100% accuracy in 10 individuals [29]. However, this study was performed in young healthy individuals with much less variable gait patterns. Furthermore, the existing FES system in this study uses a similar gait detection algorithm to many current FES systems, so the improvements made by the AFES system represent an important step forward in available technologies.

Strengths.

The main strength of this study is the customizability of the AFES system, which adjusts DF and PF stimulation amplitude based on real-time gait biomechanics. The equations used to adjust the stimulation amplitude are relatively simple linear equations, but could be easily adjusted to incorporate more complex system dynamics, which may allow the system to work for a wider range of impairment levels. Additionally, the AFES system could be adapted to stimulate different muscle groups based on selected gait parameters, making the system versatile to individual impairments and clinical conditions.

The use of the adaptive treadmill allows for real-time changes in walking speed due to changes in step length and forward propulsion [16,32]. As one goal of the AFES system is to increase forward propulsion with real-time adjustments to plantarflexor stimulation, these potential increases in propulsion can be immediately leveraged into increases in walking speed on the adaptive treadmill. While the AFES system may benefit from further development, the combination of adaptive stimulation and adaptive treadmill represents a promising new rehabilitation protocol for increasing propulsion and walking speed after stroke.

Finally, the subject pool used in this study represented a broad range of impairment levels. While this range likely washed out some effects of the stimulation at the group level, the responder versus nonresponder analysis in the discussion section presents a novel method for examining characteristics common to subjects who may benefit from stimulation therapy. With this information and additional research to further distinguish between responders and nonresponders, it may be possible to predict which subjects will respond to which treatment and optimize poststroke gait rehabilitation.

Limitations.

The main limitation of this work is that stimulating both DF and PF muscles in the same stride may lead to increased toe drag at toe-off, which is a limitation of both the FES and AFES systems. PF stimulation typically increases plantarflexion at toe-off, which can be beneficial for propulsion [10], but individuals poststroke may not be able to accommodate that increased plantarflexion and subsequently experience more toe drag before the DF stimulation is activated. Due to inherent lags in the equipment and code, the PF stimulation may have remained on for a short time into swing phase [30], causing increased plantarflexion in early swing. While the transition time between PF and DF stimulation was within recommended guidelines [56] and shorter than a few other studies [57,58], developing an improved timing algorithm presents a clear avenue for future research. A few other studies have developed predictive timing algorithms for FES during walking [24,56], and implementing a similar predictive system may reduce toe drag at toe-off.

A further limitation of this paper is the long duration of the trials, which may have contributed to muscle fatigue and altered responses to stimulation. Four-minute trials were selected to ensure that the AFES system would have enough time to detect and adapt to each subject’s gait pattern over multiple strides. All subjects were able to complete 4-min trials, as evidenced by the completion of the 6-min walk test, but steps were taken to mitigate fatigue due to multiple trials. All subjects were given time to rest between trials, and the next trial began only when the subject indicated they were ready. All subjects took no more than 5 min between any two trials, suggesting that subjects were not too tired to walk normally. Additionally, stimulation was alternated on and off in 1-min intervals [26], which reduced the total duration of walking with stimulation and may have reduced muscle fatigue. While the trial duration in this study is longer than some, it is still shorter than some training studies [13,34,59] and no subjects indicated that they could not continue due to fatigue, so fatigue is not expected to have had a significant effect on the outcomes.

Finally, while the motion capture system records kinematic data at 100 Hz, the stimulation system could not read data that quickly because of the additional time it takes to process each frame of data and calculate updated stimulation amplitudes [30]. The resultant frame rate of the AFES system was 12.5 Hz, which may reduce the precision of the calculated kinetic and kinematic gait parameters. However, this decreased frequency did not reduce the ability of the AFES system to iteratively adjust the stimulation amplitudes at each stride, as updated stimulation amplitudes were recorded for all strides. Therefore, the reduced data frequency of the AFES system is not believed to have negatively impacted the results.

Conclusion

For some subjects, the AFES system resulted in improved walking speed, dorsiflexion angle, and propulsion relative to the FES system, but other subjects did not experience benefits. In particular, PF stimulation timing was significantly improved with the AFES system. While not all subjects responded to the stimulation, the subjects who did experienced clinically meaningful improvements in functional gait biomechanics. Future work with the AFES system may result in improvements to the system, such as a predictive timing algorithm and improved footswitch detection. Additionally, future research may be able to select the optimal stimulation training paradigm for each person using baseline characteristics to result in the most clinically meaningful changes in gait. With improvements to the system and knowledge of who may benefit the most, the AFES system may be a useful rehabilitation tool. However, it is clear that further research is needed to develop a rehabilitation paradigm that works for such a heterogeneous poststroke population.

Acknowledgment

The authors would like to thank the Center for Human Research Coordination, Henry Wright, Tami Wright, Nakai Miriyoga, and Kirti Daga. This content is solely the responsibility of the authors and does not necessarily represent the official views of NIH.

Funding Data

  • National Institutes of Health National Institute of General Medical Sciences (NIH-NIGMS) (No. P20 GM103446; Funder ID: 10.13039/100000002) and the State of Delaware.

  • National Institutes of Health (NIH) (No. GM P30 103333; Funder ID: 10.13039/100000002)

  • University of Delaware Doctoral Fellowship (Funder ID: 10.13039/100006094).

Data Availability Statement

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.

Nomenclature

AFES =

adaptive functional electrical stimulation

AGRF =

anterior ground reaction force

ATM =

adaptive treadmill

DF =

dorsiflexor

FES =

functional electrical stimulation

FSTM =

fixed-speed treadmill

PF =

plantarflexor

References

1.
Benjamin
,
E. J.
,
Muntner
,
P.
,
Alonso
,
A.
,
Bittencourt
,
M. S.
,
Callaway
,
C. W.
,
Carson
,
A. P.
,
Chamberlain
,
A. M.
, et al.,
2019
, “
Heart Disease and Stroke Statistics-2019 Update: A Report From the American Heart Association
,”
Circulation
,
139
(
10
), pp.
e56
e528
.10.1161/CIR.0000000000000659
2.
Dickstein
,
R.
,
2008
, “
Rehabilitation of Gait Speed After Stroke: A Critical Review of Intervention Approaches
,”
Neurorehabil. Neural Repair
,
22
(
6
), pp.
649
660
.10.1177/1545968308315997
3.
Roelker
,
S. A.
,
Bowden
,
M. G.
,
Kautz
,
S. A.
, and
Neptune
,
R. R.
,
2019
, “
Paretic Propulsion as a Measure of Walking Performance and Functional Motor Recovery Post-Stroke: A Review
,”
Gait Posture
,
68
, pp.
6
14
.10.1016/j.gaitpost.2018.10.027
4.
Patterson
,
S. L.
,
Forrester
,
L. W.
,
Rodgers
,
M. M.
,
Ryan
,
A. S.
,
Ivey
,
F. M.
,
Sorkin
,
J. D.
, and
Macko
,
R. F.
,
2007
, “
Determinants of Walking Function After Stroke: Differences by Deficit Severity
,”
Arch. Phys. Med. Rehabil.
,
88
(
1
), pp.
115
119
.10.1016/j.apmr.2006.10.025
5.
Lynch
,
C. L.
, and
Popovic
,
M. R.
,
2008
, “
Functional Electrical Stimulation
,”
IEEE Control Syst.
,
28
(
2
), pp.
40
50
.10.1109/MCS.2007.914689.
6.
Awad
,
L. N.
,
Lewek
,
M. D.
,
Kesar
,
T. M.
,
Franz
,
J. R.
, and
Bowden
,
M. G.
,
2020
, “
These Legs Were Made for Propulsion: Advancing the Diagnosis and Treatment of Post-Stroke Propulsion Deficits
,”
J. NeuroEng. Rehabil.
,
17
(
1
), Article No. 139.10.1186/s12984-020-00747-6
7.
Bohannon
,
R. W.
,
Morton
,
M. G.
, and
Wikholm
,
J. B.
,
1991
, “
Importance of Four Variables of Walking to Patients With Stroke
,”
Int. J. Rehabil. Res.
,
14
(
3
), pp.
246
250
.10.1097/00004356-199109000-00010
8.
Jørgensen
,
H. S.
,
Nakayama
,
H.
,
Raaschou
,
H. O.
, and
Olsen
,
T. S.
,
1995
, “
Recovery of Walking Function in Stroke Patients: The Copenhagen Stroke Study
,”
Arch. Phys. Med. Rehabil.
,
76
(
1
), pp.
27
32
.10.1016/S0003-9993(95)80038-7
9.
Langhorne
,
P.
,
Coupar
,
F.
, and
Pollock
,
A.
,
2009
, “
Motor Recovery After Stroke: A Systematic Review
,”
Lancet Neurol.
,
8
(
8
), pp.
741
754
.10.1016/S1474-4422(09)70150-4
10.
Kesar
,
T. M.
,
Perumal
,
R.
,
Reisman
,
D. S.
,
Jancosko
,
A.
,
Rudolph
,
K. S.
,
Higginson
,
J. S.
, and
Binder-Macleod
,
S. A.
,
2009
, “
Functional Electrical Stimulation of Ankle Plantarflexor and Dorsiflexor Muscles: Effects on Poststroke Gait
,”
Stroke
,
40
(
12
), pp.
3821
3827
.10.1161/STROKEAHA.109.560375
11.
Li
,
S.
,
Francisco
,
G. E.
, and
Zhou
,
P.
,
2018
, “
Post-Stroke Hemiplegic Gait: New Perspective and Insights
,”
Front. Physiol.
,
9
, pp.
1
8
.10.3389/fphys.2018.01021
12.
Bowden
,
M. G.
,
Balasubramanian
,
C. K.
,
Neptune
,
R. R.
, and
Kautz
,
S. A.
,
2006
, “
Anterior-Posterior Ground Reaction Forces as a Measure of Paretic Leg Contribution in Hemiparetic Walking
,”
Stroke
,
37
(
3
), pp.
872
876
.10.1161/01.STR.0000204063.75779.8d
13.
Allen
,
J. L.
,
Ting
,
L. H.
, and
Kesar
,
T. M.
,
2018
, “
Gait Rehabilitation Using Functional Electrical Stimulation Induces Changes in Ankle Muscle Coordination in Stroke Survivors: A Preliminary Study
,”
Front. Neurol.
,
9
, p.
1127
.10.3389/fneur.2018.01127
14.
Reisman
,
D. S.
,
Kesar
,
T. M.
,
Perumal
,
R.
,
Roos
,
M.
,
Rudolph
,
K. S.
,
Higginson
,
J. S.
,
Helm
,
E.
, and
Binder-Macleod
,
S. A.
,
2013
, “
Time Course of Functional and Biomechanical Improvements During a Gait Training Intervention in Persons With Chronic Stroke
,”
J. Neurol. Phys. Ther.
,
37
(
4
), pp.
159
165
.10.1097/NPT.0000000000000020
15.
Kesar
,
T. M.
,
Reisman
,
D. S.
,
Perumal
,
R.
,
Jancosko
,
A. M.
,
Higginson
,
J. S.
,
Rudolph
,
K. S.
, and
Binder-Macleod
,
S. A.
,
2011
, “
Combined Effects of Fast Treadmill Walking and Functional Electrical Stimulation on Post-Stroke Gait
,”
Gait Posture
,
33
(
2
), pp.
309
313
.10.1016/j.gaitpost.2010.11.019
16.
Ray
,
N. T.
,
Reisman
,
D. S.
, and
Higginson
,
J. S.
,
2021
, “
Combined User-Driven Treadmill Control and Functional Electrical Stimulation Increases Walking Speeds Poststroke
,”
J. Biomech.
,
124
, p.
110480
.10.1016/j.jbiomech.2021.110480
17.
Embrey
,
D. G.
,
Holtz
,
S. L.
,
Alon
,
G.
,
Brandsma
,
B. A.
, and
McCoy
,
S. W.
,
2010
, “
Functional Electrical Stimulation to Dorsiflexors and Plantar Flexors During Gait to Improve Walking in Adults With Chronic Hemiplegia
,”
Arch. Phys. Med. Rehabil.
,
91
(
5
), pp.
687
696
.10.1016/j.apmr.2009.12.024
18.
Jiang
,
C.
,
Zheng
,
M.
,
Li
,
Y.
,
Wang
,
X.
,
Li
,
L.
, and
Song
,
R.
,
2020
, “
Iterative Adjustment of Stimulation Timing and Intensity During FES-Assisted Treadmill Walking for Patients After Stroke
,”
IEEE Trans. Neural Syst. Rehabil. Eng.
,
28
(
6
), pp.
1292
1298
.10.1109/TNSRE.2020.2986295
19.
Seel
,
T.
,
Laidig
,
D.
,
Valtin
,
M.
,
Werner
,
C.
,
Raisch
,
J. J.
, and
Schauer
,
T.
,
2014
, “
Feedback Control of Foot Eversion in the Adaptive Peroneal Stimulator
,”
2014 22nd Mediterranean Conference on Control and Automation
, Palermo, Italy, June 16–19, pp.
1482
1487
.10.1109/MED.2014.6961585
20.
Chen
,
W. L.
,
Chen
,
S. C.
,
Chen
,
C. C.
,
Chou
,
C. H.
,
Shih
,
Y. Y.
,
Chen
,
Y. L.
, and
Kuo
,
T. S.
,
2010
, “
Patient-Driven Loop Control for Ambulation Function Restoration in a Non-Invasive Functional Electrical Stimulation System
,”
Disability Rehabil.
,
32
(
1
), pp.
65
71
.10.3109/09638280903026564
21.
Seel
,
T.
,
Werner
,
C.
, and
Schauer
,
T.
,
2016
, “
The Adaptive Drop Foot Stimulator - Multivariable Learning Control of Foot Pitch and Roll Motion in Paretic Gait
,”
Med. Eng. Phys.
,
38
(
11
), pp.
1205
1213
.10.1016/j.medengphy.2016.06.009
22.
Chen
,
G.
,
Shen
,
Z.
,
Zhuang
,
Y.
,
Wang
,
X.
, and
Song
,
R.
,
2018
, “
Intensity- and Duration-Adaptive Functional Electrical Stimulation Using Fuzzy Logic Control and a Linear Model for Dropfoot Correction
,”
Front. Neurol.
,
9
, p.
165
.10.3389/fneur.2018.00165
23.
Chen
,
Y.-L.
,
Li
,
Y.-C.
,
Kuo
,
T.-S.
, and
Lai
,
J.-S.
,
2001
, “
The Development of a Closed-Loop Controlled Functional Electrical Stimulation (FES) in Gait Training
,”
J. Med. Eng. Technol.
,
25
(
2
), pp.
41
48
.10.1080/03091900110043612
24.
Li
,
Y.
,
Yang
,
X.
,
Zhou
,
Y.
,
Chen
,
J.
,
Du
,
M.
, and
Yang
,
Y.
,
2021
, “
Adaptive Stimulation Profiles Modulation for Foot Drop Correction Using Functional Electrical Stimulation: A Proof of Concept Study
,”
IEEE J. Biomed. Health Inform.
,
25
(
1
), pp.
59
68
.10.1109/JBHI.2020.2989747
25.
Melo
,
P. L.
,
Silva
,
M. T.
,
Martins
,
J. M.
, and
Newman
,
D. J.
,
2015
, “
Technical Developments of Functional Electrical Stimulation to Correct Drop Foot: Sensing, Actuation and Control Strategies
,”
Clin. Biomech.
,
30
(
2
), pp.
101
113
.10.1016/j.clinbiomech.2014.11.007
26.
Awad
,
L. N.
,
Reisman
,
D. S.
,
Kesar
,
T. M.
, and
Binder-Macleod
,
S. A.
,
2014
, “
Targeting Paretic Propulsion to Improve Poststroke Walking Function: A Preliminary Study
,”
Arch. Phys. Med. Rehabil.
,
95
(
5
), pp.
840
848
.10.1016/j.apmr.2013.12.012
27.
Conway
,
K. A.
,
Bissette
,
R. G.
, and
Franz
,
J. R.
,
2018
, “
The Functional Utilization of Propulsive Capacity During Human Walking
,”
J. Appl. Biomech.
,
34
(
6
), pp.
474
482
.10.1123/jab.2017-0389
28.
Awad
,
L. N.
,
Hsiao
,
H.
, and
Binder-Macleod
,
S. A.
,
2020
, “
Central Drive to the Paretic Ankle Plantarflexors Affects the Relationship Between Propulsion and Walking Speed After Stroke
,”
J. Neurol. Phys. Ther.
,
44
(
1
), pp.
42
48
.10.1097/NPT.0000000000000299
29.
Dong
,
H.
,
Hou
,
J.
,
Song
,
Z.
,
Xu
,
R.
,
Meng
,
L.
, and
Ming
,
D.
,
2022
, “
An Adaptive Reflexive Control Strategy for Walking Assistance System Based on Functional Electrical Stimulation
,”
Front. Neurosci.
,
16
, p.
944291
.10.3389/fnins.2022.944291
30.
Donlin
,
M. C.
, and
Higginson
,
J. S.
,
2024
, “
Adaptive Functional Electrical Stimulation Delivers Stimulation Amplitudes Based on Real-Time Gait Biomechanics
,”
ASME J. Med. Devices
,
18
(
2
), p.
021002
.10.1115/1.4065479
31.
Ray
,
N. T.
,
Reisman
,
D. S.
, and
Higginson
,
J. S.
,
2020
, “
Walking Speed Changes in Response to User-Driven Treadmill Control After Stroke
,”
J. Biomech.
,
101
, p.
109643
.10.1016/j.jbiomech.2020.109643
32.
Ray
,
N. T.
,
Knarr
,
B. A.
, and
Higginson
,
J. S.
,
2018
, “
Walking Speed Changes in Response to Novel User-Driven Treadmill Control
,”
J. Biomech.
,
78
, pp.
143
149
.10.1016/j.jbiomech.2018.07.035
33.
Kesar
,
T. M.
,
Perumal
,
R.
,
Jancosko
,
A.
,
Reisman
,
D. S.
,
Rudolph
,
K. S.
,
Higginson
,
J. S.
, and
Binder-Macleod
,
S. A.
,
2010
, “
Novel Patterns of Functional Electrical Stimulation Have an Immediate Effect on Dorsiflexor Muscle Function During Gait for People Poststroke
,”
Phys. Ther.
,
90
(
1
), pp.
55
66
.10.2522/ptj.20090140
34.
Hakansson
,
N. A.
,
Kesar
,
T. M.
,
Reisman
,
D. S.
,
Binder-Macleod
,
S. A.
, and
Higginson
,
J. S.
,
2011
, “
Effects of Fast Functional Electrical Stimulation Gait Training on Mechanical Recovery in Poststroke Gait
,”
Artif. Organs
,
35
(
3
), pp.
217
220
.10.1111/j.1525-1594.2011.01215.x
35.
Kerrigan
,
D. C.
,
Todd
,
M. K.
, and
Croce
,
U. D.
,
1998
, “
Gender Differences in Joint Biomechanics During Walking Normative Study in Young Adults
,”
Am. J. Phys. Med. Rehabil.
,
77
(
1
), pp.
2
7
.10.1097/00002060-199801000-00002
36.
Balaban
,
B.
, and
Tok
,
F.
,
2014
, “
Gait Disturbances in Patients With Stroke
,”
Phys. Med. Rehabil.
,
6
(
7
), pp.
635
642
.10.1016/j.pmrj.2013.12.017
37.
Burdett
,
R.
,
Borello-France
,
D.
,
Blatchly
,
C.
, and
Potter
,
C.
,
2016
, “
Gait Comparison of Subjects With Hemiplegia Walking Unbraced, With Ankle-Foot Orthosis, and With Air-Stirrup® Brace
,”
Phys. Ther.
,
68
(
8
), pp.
1197
1203
.10.1093/ptj/68.8.1197
38.
Olney
,
S. J.
, and
Richards
,
C.
,
1996
, “
Hemiparetic Gait Following Stroke. Part I: Characteristics
,”
Gait Posture
,
4
(
2
), pp.
136
148
.10.1016/0966-6362(96)01063-6
39.
Awad
,
L. N.
,
Kesar
,
T. M.
,
Reisman
,
D.
, and
Binder-Macleod
,
S. A.
,
2013
, “
Effects of Repeated Treadmill Testing and Electrical Stimulation on Post-Stroke Gait Kinematics
,”
Gait Posture
,
37
(
1
), pp.
67
71
.10.1016/j.gaitpost.2012.06.001
40.
McGraw
,
K. O.
, and
Wong
,
S. P.
,
1992
, “
A Common Language Effect Size Statistic
,”
Psychol. Bull.
,
111
(
2
), pp.
361
365
.10.1037/0033-2909.111.2.361
41.
Kempski
,
K. M.
,
Ray
,
N. T.
,
Knarr
,
B. A.
, and
Higginson
,
J. S.
,
2019
, “
Dynamic Structure of Variability in Joint Angles and Center of Mass Position During User-Driven Treadmill Walking
,”
Gait Posture
,
71
, pp.
241
244
.10.1016/j.gaitpost.2019.04.031
42.
Donlin
,
M. C.
,
Pariser
,
K. M.
,
Downer
,
K. E.
, and
Higginson
,
J. S.
,
2022
, “
Adaptive Treadmill Walking Encourages Persistent Propulsion
,”
Gait Posture
,
93
, pp.
246
251
.10.1016/j.gaitpost.2022.02.017
43.
Wonsetler
,
E. C.
, and
Bowden
,
M. G.
,
2017
, “
A Systematic Review of Mechanisms of Gait Speed Change Post-Stroke. Part 1: Spatiotemporal Parameters and Asymmetry Ratios
,”
Top. Stroke Rehabil.
,
24
(
6
), pp.
435
446
.10.1080/10749357.2017.1285746
44.
Hsiao
,
H.
,
Knarr
,
B. A.
,
Higginson
,
J. S.
, and
Binder-Macleod
,
S. A.
,
2015
, “
Mechanisms to Increase Propulsive Force for Individuals Poststroke
,”
J. NeuroEng. Rehabil.
,
12
(
1
), Article No. 40.10.1186/s12984-015-0030-8
45.
Perry
,
J.
,
Garrett
,
M.
,
Gronley
,
J. K.
, and
Mulroy
,
S. J.
,
1995
, “
Classification of Walking Handicap in the Stroke Population
,”
Stroke
,
26
(
6
), pp.
982
989
.10.1161/01.STR.26.6.982
46.
Lewek
,
M. D.
, and
Sykes
,
R.
,
2019
, “
Minimal Detectable Change for Gait Speed Depends on Baseline Speed in Individuals With Chronic Stroke
,”
J. Neurol. Phys. Ther.
,
43
(
2
), pp.
122
127
.10.1097/NPT.0000000000000257
47.
van de Port
,
I. G.
,
Kwakkel
,
G.
, and
Lindeman
,
E.
,
2008
, “
Community Ambulation in Patients With Chronic Stroke: How Is It Related to Gait Speed?
,”
J. Rehabil. Med.
,
40
(
1
), pp.
23
27
.10.2340/16501977-0114
48.
Patterson
,
K. K.
,
Parafianowicz
,
I.
,
Danells
,
C. J.
,
Closson
,
V.
,
Verrier
,
M. C.
,
Staines
,
W. R.
,
Black
,
S. E.
, and
McIlroy
,
W. E.
,
2008
, “
Gait Asymmetry in Community-Ambulating Stroke Survivors
,”
Arch. Phys. Med. Rehabil.
,
89
(
2
), pp.
304
310
.10.1016/j.apmr.2007.08.142
49.
Kesar
,
T. M.
,
Binder-Macleod
,
S. A.
,
Hicks
,
G. E.
, and
Reisman
,
D. S.
,
2011
, “
Minimal Detectable Change for Gait Variables Collected During Treadmill Walking in Individuals Post-Stroke
,”
Gait Posture
,
33
(
2
), pp.
314
317
.10.1016/j.gaitpost.2010.11.024
50.
Tilson
,
J. K.
,
Sullivan
,
K. J.
,
Cen
,
S. Y.
,
Rose
,
D. K.
,
Koradia
,
C. H.
,
Azen
,
S. P.
, and
Duncan
,
P. W.
, and
for the Locomotor Experience Applied Post Stroke (LEAPS) Investigative Team
,
2010
, “
Meaningful Gait Speed Improvement During the First 60 Days Poststroke: Minimal Clinically Important Difference
,”
Phys. Ther.
,
90
(
2
), pp.
196
208
.10.2522/ptj.20090079
51.
Chui
,
K.
,
Hood
,
E.
, and
Klima
,
D.
,
2012
, “
Meaningful Change in Walking Speed
,”
Top. Geriatr. Rehabil.
,
28
(
2
), pp.
97
103
.10.1097/TGR.0b013e3182510195
52.
Bohannon
,
R. W.
, and
Glenney
,
S. S.
,
2014
, “
Minimal Clinically Important Difference for Change in Comfortable Gait Speed of Adults With Pathology: A Systematic Review
,”
J. Eval. Clin. Pract.
,
20
(
4
), pp.
295
300
.10.1111/jep.12158
53.
Burke
,
E.
,
Dodakian
,
L.
,
See
,
J.
,
McKenzie
,
A.
,
Riley
,
J. D.
,
Le
,
V.
, and
Cramer
,
S. C.
,
2014
, “
A Multimodal Approach to Understanding Motor Impairment and Disability After Stroke
,”
J. Neurol.
,
261
(
6
), pp.
1178
1186
.10.1007/s00415-014-7341-8
54.
Hsiao
,
H.
,
Knarr
,
B. A.
,
Higginson
,
J. S.
, and
Binder-Macleod
,
S. A.
,
2015
, “
The Relative Contribution of Ankle Moment and Trailing Limb Angle to Propulsive Force During Gait
,”
Human Mov. Sci.
,
39
, pp.
212
221
.10.1016/j.humov.2014.11.008
55.
Kepple
,
T. M.
,
Siegel
,
K. L.
, and
Stanhope
,
S. J.
,
1997
, “
Relative Contributions of the Lower Extremity Joint Moments to Forward Progression and Support During Gait
,”
Gait Posture
,
6
(
1
), pp.
1
8
.10.1016/S0966-6362(96)01094-6
56.
Parthasarathy
,
A.
,
V.n
,
M.
, and
Talasila
,
V.
,
2020
, “
Forecasting a Gait Cycle Parameter Region to Enable Optimal FES Triggering
,”
IFAC-PapersOnLine
,
53
(
1
), pp.
232
239
.10.1016/j.ifacol.2020.06.040
57.
Chen
,
G.
,
Ma
,
L.
,
Song
,
R.
,
Li
,
L.
,
Wang
,
X.
, and
Tong
,
K.
,
2018
, “
Speed-Adaptive Control of Functional Electrical Stimulation for Dropfoot Correction
,”
J. NeuroEng. Rehabil.
,
15
(
1
), Article No. 98.10.1186/s12984-018-0448-x
58.
Skelly
,
M. M.
, and
Chizeck
,
H. J.
,
2001
, “
Real-Time Gait Event Detection for Paraplegic FES Walking
,”
IEEE Trans. Neural Syst. Rehabil. Eng.
,
9
(
1
), pp.
59
68
.10.1109/7333.918277
59.
Awad
,
L. N.
,
Reisman
,
D. S.
,
Wright
,
T. R.
,
Roos
,
M. A.
, and
Binder-Macleod
,
S. A.
,
2014
, “
Maximum Walking Speed Is a Key Determinant of Long Distance Walking Function After Stroke
,”
Top. Stroke Rehabil.
,
21
(
6
), pp.
502
509
.10.1310/tsr2106-502

Supplementary data