Augmenting gait in a population exhibiting foot drop with adaptive functional electrical stimulation

Participant recruitment

A total of 60 participants were assessed for eligibility. Study inclusion criteria specified participants must be an adult aged 18-70 living with a lower extremity impairment that makes walking difficult or uncomfortable; be capable of sitting, standing, and walking independently or with assistance; be able to walk at least 50 feet independently or with assistance; and be able to understand and follow basic instructions in English. Of those assessed, eight were excluded due to not meeting eligibility requirements, declining to participate, or other. Fifty-two were enrolled in the study. Of the enrollees, fourteen were lost to follow up due to scheduling logistics. Thirty-eight were allocated for data collection. Four of the allocated were excluded for insufficient endurance to complete the experimental protocol, and two were excluded due to equipment errors. The final analysis pool contained 32 subjects.

Ethical considerations

The present study was carried out according to the Declaration of Helsinki. Participation was voluntary and based on written informed consent from the participant. The study has been approved by the Institutional Review Board at Ethical & Independent Review Services (E&I 21039-01A) and the Institutional Review Board at Cleveland State University (IRB-FY2021-301).

Study design

The study design is a one-group, pre-test post-test study where we assessed the impact of FES relative to each participant’s unstimulated gait in a pairwise fashion. All participants received adaptive FES to the dorsiflexors and evertors during up to three back-to-back walking sets.

Technology

In this study, we use the Cionic Neural Sleeve, an adaptive, current-steering FES device to stimulate dorsiflexors and evertors to improve ankle movement. The novel technology includes an array of FES gel electrodes placed on the anterolateral region of the impaired lower leg (Figure 2), corresponding with the atlas of the muscle motor points for the lower limb²¹. The frequency (Hz) and pulse width (µs) of all electrodes are tunable. Additionally, the current intensity (mA) of each electrode in the array can be independently configured using a mobile application. This multi-channel current-steering allows for precision control of ankle movement in multiple planes. The sequencing of muscle stimulation is controlled by algorithms operating on data streamed from on-body sensors. Specifically, the algorithms use spatial orientation data to trigger muscle contraction and relaxation events that adapt to changes in gait speed.

• Download figure
• Open in new tab

Figure 1.

Flow chart on participant inclusion. Of 60 assessed for eligibility, 52 were enrolled. Thirty-eight of the enrollees were allocated for data collection, and 32 were included in the final pool. Reasons for exclusion from analysis include inability to complete protocol and equipment errors.

• Download figure
• Open in new tab

Figure 2.

Left: Cionic Neural Sleeve showing the configuration of the FES and EMG electrode arrays. Center: Cionic mobile application interface for setting FES parameters (frequency, pulsewidth, intensity) and current field steering. Right: Cionic Neural Sleeve on body.

Data collection

Prior to walking with FES, each participant underwent a standard seated FES configuration protocol facilitated by trained researchers to set stimulation parameters to achieve observable dorsiflexion (ideally exceeding 10°) with balanced eversion while maintaining participant comfort. Frequency and pulse width parameters were set to 40 Hz and 300 µs, respectively, and intensity was increased until the desired ankle movement was achieved. Frequency and pulse width could also be tuned in increments of 5 Hz and 100 µs as needed for comfort or efficacy. FES configuration was completed using the Cionic mobile application. Each participant gained familiarity with stimulated gait by traversing a 40-foot walkway as a warmup. They walked with their typical assistive devices, if applicable, excluding AFOs.

Next, each participant alternated between unstimulated (noFES) walking collection of their typical gait and stimulated (FES) walking collection. Data was collected from the Cionic Neural Sleeve using the following sensors: Bosch Sensortec inertial measurement units (IMUs) placed on the thigh, shank, and foot to reconstruct the leg’s spatial orientation and three foot pressure sensor pads sampled using an analog-to-digital converter (ADS1298CZXGT, Texas Instruments) in the shoe for gait event detection. IMUs were calibrated prior to each participant’s collections. For the stimulated collections, the dorsiflexors were stimulated to maintain toe clearance during swing phase and improve heel strike, and evertors to achieve neutral ankle at heel strike and loading response. Unstimulated and stimulated collections were performed back-to-back up to three times (six total collections) to directly quantify the impact of FES. Each collection consisted of walking across the walkway, turning around, and walking back to the starting point. Time intervals corresponding to the turn around were identified via manual inspection of kinematics data and were removed from the analysis. The first two steps and last step for each length of the walkway were removed from the analysis. Pressure sensor data was automatically labeled and manually inspected to extract heel strike, toe strike, and toe off events.

Statistical analyses

Based on our statistical power analysis, we required eighteen or more participants to achieve 80% power to reject the null hypothesis which states there is no difference between pre-test (noFES) and post-test (FES) conditions in a pairwise comparison (α = 0.05). We selected an effect size of 0.7 for the power analysis based on a previous study of FES impact on post-stroke participants²².

In this analysis, we examined whether the impact of FES on gait is statistically significant relative to unstimulated gait by pairwise comparison. For each metric, we averaged each participant’s metric values across steps for both noFES and FES walking. Each metric’s final analysis pool contained 32 noFES and 32 FES samples. The Shapiro-Wilk test for normality²³ was used to assess the assumption of normality and paired t-tests were used to statistically compare noFES and FES foot drop metrics if the assumption for normality was met (all metrics except foot angle at heel strike). If the normality assumption was not met, the Wilcoxon signed rank test was used as an alternative to the paired t-test (this was the case for foot angle at heel strike). Bonferroni multiple hypothesis correction was performed (four metrics tested), and our corrected significance level for comparison was 0.0125 (0.05 divided by the four hypotheses tested). We also report the 95% confidence interval for the differences between FES and noFES foot drop metrics.

Participant demographics

The final analysis pool was split across 32 participants aged 61.1±10.9 living near San Francisco and Cleveland. The subjects included 18 females and 14 males. All subjects had a diagnosis of neurological disease or injury: 20 were post-stroke, eight had multiple sclerosis, two had a spinal cord injury, one had cerebral palsy, and one had idiopathic foot drop. All exhibited symptoms of foot drop. The average stimulation parameters across participants included an amplitude of 44±11 mA, frequency of 38±14 Hz, and a pulse duration of 306±51 µs.

Primary outcomes

The impacts of FES on dorsiflexion across 32 participants exhibiting symptoms of foot drop are shown in Figure 3A. The kinematics profile shows average dorsiflexion throughout the gait cycle for all participants, and the shaded region shows the standard deviation. Zero percent corresponds to heel strike. Additionally, the participants’ baseline noFES kinematics showing symptoms of foot drop are in orange while Cionic FES show the impact of gait assist in blue. We observed that the primary impact of FES on dorsiflexion occurs at heel strike and during swing phase (approximately to 65-100% of gait), and that dorsiflexion at heel strike is increased by an average of 5.2° (CI[3.5°, 6.8°]) with statistical significance (p-value=7e-5). The average increase of 5.2° is greater than the minimal clinically important difference of 4.9° as defined by the Clinical Practice Guideline for the Use of Ankle-Foot Orthoses and Functional Electrical Stimulation Post Stroke²⁴. All but one participant showed increased dorsiflexion at heel strike; the biggest observed increase from a participant was an average of 14.6°, while one decreased by -1.6°.

• Download figure
• Open in new tab

Figure 3.

(A) Ankle dorsiflexion gait profiles averaged over the 32 participants walking with Cionic FES (blue profile) and without (orange profile). The gait cycle profile is represented as a percent, where 0% occurs at heel strike. The solid lines represent the cohort average, and the shaded intervals are ± one standard deviation. (B) Ankle inversion gait profiles averaged over the 32 participants with Cionic FES (blue) and without (orange). (C) Stacked degrees of dorsiflexion improvement (gray bar) and inversion improvement (blue bar) for each participant. All participants but one showed a net improvement across both primary outcomes. *The gray normative profile was determined from the typical walking gait of eleven normative participants. This was a sample of convenience conducted on Cionic employees with no known gait impairments. Normative participant data was collected under IRB oversight and consisted of 11 participants (10 male, 1 female) with age range [24-53].

In addition to decreased dorsiflexion, many of the 32 participants also exhibit excessive ankle inversion as shown in Figure 3B. Excessive inversion is especially apparent for noFES during the swing phase. Evertor FES is configured to reduce ankle inversion during swing phase through loading response (Figure 3B) as this increases lateral loading response stability. We show an improved outcome by reducing mean inversion during the swing phase by an average of -3.6° (CI[-5.3°, -1.8°], p-value=3e-4). The largest observed decrease was an average of -21.6°. In total 88% of participants showed a positive improvement to ankle inversion during swing.

Figure 3C shows the stacked degrees of improvement for dorsiflexion and inversion for each participant, providing a more comprehensive picture of the impact of gait assist across multiple movements. Some participants exhibit large improvements to inversion due to high baseline inversion, while others show higher dorsiflexion improvement relative to inversion. All participants but three showed an improvement in both dorsiflexion and inversion, and all but one showed net improvement across both primary metric outcomes.

Secondary outcomes

We observed an average of 5.5° (CI[3.9°, 7.0°]) increase in foot angle at heel strike (p-value=9e-7). Heel-toe time increased by an average of 3.4% of gait (CI[1.9%, 4.8%], p-value=4e-5). After FES treatment, average foot angle at heel strike and heel-toe time were closer to the normative average. With dorsiflexor stimulation, foot angle increased in all participants except one, whose foot angle decreased by -0.4°. The largest FES-driven increase in foot angle was 15.8°. Additionally, stimulated gait resulted in increased heel-toe time in all but three participants, who decreased by nor more than 1.8%. The participant with the greatest increase in heel-toe time showed a 17.5% increase.

Study Limitations

This study did not include the collection of certain traditional gait analysis metrics including spatial measures like gait speed and stride length as we relied solely on the on-body device for all data collection and metric extraction. Additionally, we were not able to compare FES treatment impacts relative to standard of care, i.e. AFO, in a randomized control trial. An RCT remains a future research study.