Metabolic energy expenditure is a key measure for locomotor performance when assessing wearable robotic devices. Individuals with transtibial amputation generally have greater metabolic energy expenditure walking with a passive lower limb prosthesis compared with able-bodied individuals due to the lack of a biological ankle joint.1–3 Passive energy restoring feet and powered prostheses can reduce metabolic energy expenditure compared with traditional passive prostheses.4–7 However, there are mixed results in the literature on the metabolic benefit of powered prostheses, potentially due to the user’s functional classification. Higher functioning K4 ambulators and active-duty military showed more metabolic benefits than a lower functional class of K3.8,9 Typically, metabolic energy expenditure has been evaluated on a treadmill, which does not necessarily capture how individuals with amputation would walk with a prosthesis in real-world environments. One study by Au et al4 evaluated the outdoor metabolic cost of transporting an experimental powered prosthesis while walking around an outdoor track. An outdoor track is a valuable preliminary step but still does not fully simulate all aspects of a real-world environment, such as obstacle avoidance, changes in direction, and uneven sidewalks.
One factor limiting the performance of a powered lower limb prosthesis is the controller. Commercially available powered prostheses can provide net positive mechanical work at the end of the stance phase when using state-machine controllers sensing kinematics and/or kinetics.4,5,8 Outside the laboratory, commercially available powered prostheses may not enhance mobility in real-world use.10 Controllers that include machine learning algorithms can allow for increased variability in walking patterns,11–13 but reliance on gait or terrain classification approaches do not typically provide the ability to navigate crowds, obstacles, and volitional changes in stepping. Electromyography (EMG) is one way to provide a volitional prosthetic input to provide greater flexibility to the prosthetic controller for complex real-world environments. Direct EMG control is a subset of EMG controllers where the EMG signal directly modulates prosthetic dynamics.14 Individuals with amputation can learn to activate their residual limb muscle activity volitionally to control a prosthesis for navigating slopes and stairs.15–17 Prosthetic controllers are often evaluated in laboratory settings with ramp and stair configurations, or, even obstacle courses,18–20 but this does not capture all the complexities of the natural environments encountered in daily life. In order to fully evaluate prosthetic controllers, investigations outside of the laboratory are needed to better evaluate controller and prosthetic performance.
The goal of this study was to measure the metabolic cost of transport for individuals with amputation walking outdoors with a portable bionic ankle under proportional myoelectric control compared with a prescribed passive prosthesis. We had 6 participants complete a predetermined outdoor loop walking with their prescribed passive prosthesis and a powered prosthesis under myoelectric control. We hypothesized that walking with a powered prosthesis under continuous myoelectric control would reduce the metabolic cost of transport compared with walking with a passive prosthesis.
Methods
We recruited 6 participants with a unilateral transtibial amputation to walk an outdoor course. Participant demographics are provided in Table 1. All participants provided informed and written consent to a protocol previously approved by the University of Florida Institutional Review Board. We verified all participants’ ability to volitionally contract their residual muscles via manual and EMG testing. Individuals with amputation due to trauma or vascular disease were recruited if they were capable of walking over 10 minutes continuously without a mobility aid and had volitional control of their residual gastrocnemius. We based the sample size for the study on prior data from our laboratory evaluating myoelectric control of pneumatic prosthesis during treadmill walking,21 as well as a study by Au et al4 reporting significant reductions in cost of transport in 3 participants using a powered prostheses walking on an outdoor track.
Participant Demographics
| Participant number | Gender | Age, y | Height, m | Weight, kg | Years postsurgery | Reason for amputation |
|---|---|---|---|---|---|---|
| 1 | M | 54 | 1.75 | 86 | 1.5 | Vascular |
| 2 | F | 48 | 1.73 | 87 | 2.0 | Vascular |
| 3 | M | 66 | 1.80 | 92 | 0.4 | Vascular |
| 4 | F | 17 | 1.63 | 64 | 2.0 | Trauma |
| 5 | M | 65 | 1.95 | 93 | 38.5 | Trauma |
| 6 | F | 23 | 1.57 | 91 | 0.5 | Cancer |
| Mean (SD) | 45.5 (20.9) | 1.73 (0.13) | 85.5 (10.9) | 7.5 (15.2) |
Note: The final row displays the mean and 1 SD.
Prosthesis and Controller
For this study, we implemented a myoelectric controller on the Open Source Leg (OSL; version 1) ankle developed at the University of Michigan.22,23 The OSL, with a 30° range of ankle motion from 10° dorsiflexion to 20° plantar flexion, was equipped with a 6-axis load cell M3564F (Sunrise Instruments) and weighed 2.25 kg. The myoelectric controller ran on a Raspberry Pi 4 in Python at 450 Hz. EMG was collected via the Coapt amplifier (Coapt, LLC) using 1.3 mm Neuroline Surface Electrodes 715 (Ambu). A front pack worn by each participant weighed 1.4 kg, containing the Raspberry Pi, sync generator, batteries (Venom Fly 30 C/11.1 V), and EMG system.
The controller design was impedance-based where residual gastrocnemius muscle activity influenced the set angle of the OSL.24 We placed EMG sensors on the residual gastrocnemius and tibialis anterior muscle, even though only the gastrocnemius was used for control. Processed and smoothed real-time EMG generated a continuous proportional myoelectric control signal. The residual limb gastrocnemius EMG signal was high-pass filtered (second-order Butterworth 70 Hz) to attenuate signal artifacts, then full wave rectified the high-passed signal and thereafter low-pass filtered (second-order Butterworth 3 Hz).25–27 We determined a maximum voluntary contraction during each session for participant’s residual gastrocnemius to standardize muscle activity for the prosthetic controller input.28–31 A percentage (between 50% and 75%) of the participant’s maximum voluntary contraction normalized the EMG signal. The normalized continuous control signal mapped EMG to impedance set angle where a normalized value of 1 corresponded to maximum plantar flexion (30°). When the participants’ muscle activity was below a baseline threshold, the prosthesis held between a 4° to 6° dorsiflexed position. A slightly more dorsiflexed position was chosen to provide more natural ankle rollover during walking as well as facilitate easier balance during standing. The controller switches between 3 sets of impedance parameters. The stiffness (K) and damping (B) values are based on the stiffness and damping values for Dephy’s impedance controller since the OSL motor is a Dephy Actpack. For swing and early stance, stiffness is set to K = 2000 and damping B = 5500. As participants began to load the prosthesis, we reduced the stiffness and damping parameters to more closely mimic a biological ankle impedance behavior. This transition was determined by the sagittal plane moment of the onboard load cell. We again used the sagittal plane moment combined with the load on the prosthesis. Late stance and the transition to plantar flexion push-off was determined by the sagittal plane moment and when the load was greater than 65% of a participant’s body weight. For plantar flexion, K = 2000 and B is between 5777 and 7722 proportional to participant’s body weight ranging from 63.5 to 95.25 kg. Once the load of the prosthesis was less than 30% of the participant body weight, the parameters transitioned back to early stance/swing values. At all points during the gait cycle, EMG controlled the set angle of the impedance controller. The OSL also uses the carbon fiber Ossur Pro-Flex foot that provides additional compliance.
Data Collection
A certified prosthetist determined proper socket alignment for each participant (Table 2). Due to the height of the OSL, shorter participants or participants with longer residual limbs needed length discrepancy corrections. The prosthetist fit 3 participants (2, 4, and 6) with an EVENup to correct intact leg length discrepancies, while other participants had a heel wedge in their intact limb shoe. Additionally, participants 2, 4, and 6 had a slightly larger shoe on their amputated size because the smallest shoe size compatible with the OSL was a women’s size US 8.
Participant Prosthetic Components
| Participant number | Suspension type | Prosthetic liner | Prescribed prosthesis | Shoe size US | Open Source Leg setup |
|---|---|---|---|---|---|
| 1 | Pin | Gel | Kinterra Proteor | 10 M | No additional offsets |
| 2 | Suction | Gel | Ossur Talux | 6 W | EVENup |
| 3 | Pin | Gel | Ottobock Trias | 10 M | No additional offsets |
| 4 | Pin | Gel | Ossur Vari-Flex | 5 W | EVENup |
| 5 | Suction | Gel | Elite Blade Endolite | 13 M | Heel Wedge |
| 6 | Pin | Gel | Kinterra Proteor | 10 M | No additional offsets |
Each participant walked with the prosthesis under myoelectric control and their own prescribed passive prosthesis over a predetermined outdoor course on the University of Florida campus. The prescribed passive prostheses were all considered energy-storing and return types. Participants completed 2 loops of an outdoor course on the University of Florida campus walking with the powered prosthesis under myoelectric controller and their passive prosthesis. Additionally, each participant completed the outdoor testing loop of 0.45 km with their passive prosthesis and the powered prosthesis during a previous session for familiarization and determining walking pace (Figure 1). Metabolic data collections were done in the morning on the same day, and we asked the participants to refrain from eating and avoid caffeine and nicotine the night before and prior exercise in order to reduce variability in metabolic measures. Participants stood for 6 minutes, and an average of the last 2 minutes was used to determine their standing baseline metabolic cost. To determine the net metabolic cost, we subtracted the standing baseline from each measure for each prosthesis condition. The prosthesis condition order was randomized, and participants were given rest between each lap to return to baseline metabolic cost before beginning the next trial. Walking pace was determined by the slowest self-selected walking speed of the participant using their passive prosthesis and the powered prosthesis during a previous session. During the outdoor laps, a researcher verbally provided pace feedback based on a GPS smartwatch to control walking speed. All participants had at least 3 hours of walking inside and outside with the powered prosthesis prior to this collection. Individual familiarization time is in Supplementary Table S1 (see Supplementary Materials [available online]).
(A) Participant walking with the Open Source Leg under myoelectric control with the metabolic system on the outdoor loop. (B) Map of the outdoor course located on the University of Florida campus. One loop was 0.45 km. The dot indicates the beginning and end of a lap.
Citation: Journal of Applied Biomechanics 2025; 10.1123/jab.2024-0081
Study Measures
Surface EMG was recorded from the thigh and shank muscles for both legs, as well as respiratory measurements with a portable metabolic system while participants walked outside. Cometa MiniWave EMG (Cometa, sampled at 2000 Hz) measured muscle activity of the rectus femoris, vastus lateralis, semitendinosus, and biceps femoris from both limbs. Intact limb tibialis anterior and medial and lateral gastrocnemius were measured at 450 Hz through the Coapt amplifier and Raspberry Pi. Thigh and intact shank EMG were measured with round Ag/AgCl Kendall H124SG electrodes (diameter: 23 mm) and approximate interelectrode distance of 24 mm. Electrode placements followed the surface EMG for noninvasive assessment of muscles recommendations.32 We recorded either the residual medial or lateral gastrocnemius and tibialis anterior muscle activity at 450 Hz with round Neuroline electrodes (Ag/AgCl, 30 × 22 mm) with interelectrode distance approximately 28 mm. Palpation techniques based on previous literature determined residual limb electrode placement.33 EMG prep consisted of shaving the area (if hair was present) and cleaning the skin surface with an alcohol wipe. Inertial measurement units (Cometa Wave Track inertial system) placed on the feet measured heel strikes to parse EMG measures by stride.34–36 We measured oxygen consumption and carbon dioxide production via a COSMED K5 system (Rome, Italy). The metabolic system added 0.9 kg to the participants. Additionally, the COSMED recorded GPS speed and heart rate data obtained from a chest strap monitor that was time synced to the metabolic data.
Data Analysis
We analyzed metabolic cost of transport to evaluate prosthesis outdoor walking. The outdoor metabolic cost (Watts per kilogram) was calculated via the Brockway equation using VO2 and VCO2 and filtered with a fourth-order, low-pass Butterworth filter (cutoff frequency of 0.1 Hz).37 Cost of transport was calculated by dividing the metabolic cost by walking speed. We analyzed data from 1 powered lap and 1 passive lap for each participant because we did not have 2 continuous walking laps for all participants in the powered prosthesis condition. Additionally, the metabolic measures only contain 5 participants, since participant 1 experienced technical issues with the powered prosthesis and did not have a continuous lap with the powered prosthesis. We normalized course completion from 0% to 100% based on lap time and present the time series of the metabolic cost since steady state was not achievable during the outdoor course.38,39 We calculated the average cost of transport using 20% to 100% of a lap to account for the expected rise in oxygen uptake that occurs at the onset of walking. Data from a GPS smart watch from each lap generated an average course elevation profile. Based on the elevation profile, we segmented the course into uphill (38%–55% of course) and downhill (58.5%–81.5% of course) portions. A linear fit of the elevation profile determined the gradient of each course completion (uphill 0.006% grade and downhill 0.014% grade). We then computed an average cost of transport for the uphill and downhill section.
We calculated step-normalized EMG root mean squared (RMS) to understand differences in muscle activity during outdoor walking between our powered prosthesis and participants’ passive devices. We selected 15 strides from the flattest section of the course (25%–35% course completion) for analysis. All EMG was resampled to 450 Hz and high-pass filtered (zero-lag second-order Butterworth 50 Hz). Then, the high-pass signal was full wave rectified, and the rectified signal was low-pass filtered (zero-lag second-order Butterworth 10 Hz). We identified strides using heel strikes determined by inertial measurement units on the feet to normalize and calculate metrics per stride.35,36 The average root mean square value normalized the EMG for each muscle. To evaluate muscle activation profiles, we calculated EMG RMS during a stride for thigh and shank EMG of both limbs. Complete EMG data was only available for participants 1 through 5 due to loss of signal quality in participant 6.
After completing all walking conditions, participants completed a preference survey.10 The preference questionnaire was on a 10 cm visual analog scale from passive prosthesis (0) to myoelectric control (10). A score of 5 would indicate no preference toward either prosthesis. The questionnaire asked 1 question about which prosthesis participants preferred when walking outside.
Statistical Analysis
We analyzed outdoor EMG and metabolic measures comparing the powered and passive prostheses with paired t tests unless measures violated assumptions of normality based on Shapiro–Wilk test. Analysis for downhill cost of transport, walking speed, and RMS EMG were performed with paired t tests with SPSS (version 29.0). An alpha level of .05 determined statistical significance. The RMS of the intact limb biceps femoris and lateral gastrocnemius and uphill cost of transport violated normality, so we used a paired Wilcoxon rank test. We assessed if prosthesis preference was significantly different from 5 cm (score of no preference) using a 1 sample t test for evaluating preference for outdoor walking. Hedges g were calculated for dependent variables of interest from SPSS. For nonparametric tests, an effect size of r was reported.40 Effect sizes >0.2, >0.5, and >0.8 were considered to be of a small, moderate, and large magnitude, respectively.
Results
There were no significant differences in metabolic measures walking outside between the powered prosthesis under myoelectric control and the participants’ passive prosthesis. Participants walked at similar speeds using both prostheses (P = .266, g = 0.244) with an average and standard deviation of 0.91 (0.09) m·s−1 and 0.92 (0.08) m·s−1 using the powered and passive prostheses, respectively (Figure 2). For heart rate, there were no significant differences (P = .395, g = 0.141) when participants walked with the powered prosthesis (123 [24] beats·min−1) or their passive device (128 [23] beats·min−1). Based on an average for each course lap, we also saw no significant differences for the cost of transport (P = .142, g = 0.554) using the powered (4.80 [1.04] J·kg−1·m−1), and the passive prosthesis (4.30 [0.52] J·kg−1·m−1). Figure 2 displays the cost of transport and elevation profile over the outdoor course. Similarly, there were no significant differences for inclined walking (P = .138, r = .47) with the prosthesis under myoelectric control (5.68 [1.35] J·kg−1·m−1) and the participant’s passive device (4.80 [0.62] J·kg−1·m−1) for the inclined section of the course. The downhill cost of transport between the powered (4.75 [0.94] J·kg−1·m−1) and passive (4.07 [0.51] J·kg−1·m−1) cost of transport was not statistically different from each other, but there was a medium magnitude of effect size (P = .125, g = 0.691). Figure 3 shows the uphill and downhill cost of transport. See Supplementary Tables S2 through S7 in Supplementary Materials (available online), which contain individual participant data.
Cost of transport (J·kg−1·m−1), average speed (m·s−1), and gradient (%) of walking on the outdoor course over the total course distance. Blue represents the participant’s passive prosthesis, while orange is the powered prosthesis under myoelectric control. Vertical dotted lines and dashed lines represent the uphill and downhill segments of the course, respectively. There were no significant differences in cost of transport and walking speed between the 2 conditions. Shaded regions represent ±1SD. Metabolic data is for participants 2 through 6. There were no significant differences in cost of transport between prostheses. (Color figure online)
Citation: Journal of Applied Biomechanics 2025; 10.1123/jab.2024-0081
Bar plot showing the net cost of transport (J·kg−1·m−1) for the downhill (58.5%–81.5%) and uphill (38%–55%) sections of the outdoor course. The average cost of transport was highest for the powered prosthesis under myoelectric control (orange) compared with the participant’s passive prosthesis (blue) for uphill and downhill walking. The differences were not significant. The mean walking speed (SD) for the uphill and downhill were 0.8 (0.1) m·s−1, 0.8 (0.1) m·s−1 and 0.9 (0.1) m·s−1, 1.0 (0.1) m·s−1 for the powered and passive prostheses, respectively. Individual dots represent individual subjects. The bars represent ±1 SD. Metabolic data is for participants 2 through 6. There were no significant differences in cost of transport for uphill or downhill walking between prostheses. (Color figure online)
Citation: Journal of Applied Biomechanics 2025; 10.1123/jab.2024-0081
Walking with the powered prosthesis had greater residual limb quadriceps muscle activation compared with walking with the passive prosthesis. The time series of the residual thigh and shank EMG is presented in Figure 4. For the residual limb vastus lateralis, there were significant differences between conditions and a large effect size in muscle activity normalized by the average passive RMS EMG value for participants walking (P = .042, g = 1.056) with their passive prosthesis (0.72 [0.62]) and the prosthesis under myoelectric control (1.00 [0.87]). Similarly, there were significant differences and a large effect size in residual rectus femoris activity (P = .029, g = 1.192) between passive (0.92 [2.27]) and powered (1.29 [3.19]) prosthesis conditions during walking. Residual gastrocnemius muscle EMG RMS was not significantly different between the passive (0.63 [0.70]) and powered (1.66 [2.12]) prosthesis but trended toward an increase in RMS EMG and a large effect size during outdoor walking with the powered prosthesis (P = .062, g = 0.917). We did not see any significant differences in tibialis anterior RMS muscle activity for participants walking with their passive (0.63 [0.81]) and our powered (0.15 [3.03]) prosthesis (P = .394, g = 0.341). Additionally, there were no significant differences in normalized RMS muscle for the residual limb biceps femoris (P = .187, g = 0.377) or semitendinosus (P = .591, g = 0.208) muscles. The mean (SD) of normalized RMS EMG of the semitendinosus (0.66 [0.71]; 0.61 [1.43]) and biceps femoris (0.74 [0.54]; 0.74 [0.54]) for passive and powered walking, respectively, were similar.
Stride normalized rectified EMG of the TA, GAS, VL, RF, BF, and SM muscles on the prosthetic (left) and intact (right) limb for participants walking with the powered and their passive prostheses during outdoor walking. For the intact limb, the medial and lateral gastrocnemius were recorded. Blue represents the participant’s passive prosthesis while orange is the powered prosthesis under myoelectric control for average of 15 strides during the flat section of the course (25%–35%). For the intact limb gastrocnemius muscles, blue and green represent the passive conditions, and orange and red are the powered conditions for the medial and lateral gastrocnemius, respectively. EMG is normalized to the average root mean square of rectified muscle activity during passive walking. Heel strike represents 0% on the x-axis and shaded regions represent ±1SD. Note the results for residual biceps femoris is for participants 1, 2, 3, and 5 because participant 4’s electrode came off partway through the powered walking condition. Walking with the powered compared with the passive prostheses, participants increased their residual limb muscle activity primarily during early to midstance for quad and hamstring muscles. Participants also increased their residual gastrocnemius activity to time powered prosthesis plantar flexion. BF indicates biceps femoris; EMG, electromyography; GAS, gastrocnemius; RF, rectus femoris; SM, semitendinosus; TA, tibialis anterior; VL, vastus lateralis. (Color figure online)
Citation: Journal of Applied Biomechanics 2025; 10.1123/jab.2024-0081
There were significant differences in semitendinosus muscle activity within the intact limb between the 2 prosthesis conditions. The time series of the intact thigh and shank EMG is shown in Figure 4. There were significant differences and a large effect size in intact semitendinosus muscle activity (P = .031, g = 1.163) with increased muscle activity walking with the powered prosthesis compared with the passive prosthesis, respectively (0.99 [0.79]; 0.61 [0.60]). For intact shank muscle activity, there were no significant differences in the lateral (P = .500; r = .213) or medial gastrocnemius (P = .675, g = 0.161), respectively, for passive (0.70 [0.61]; 0.62 [0.59]) compared with powered prosthesis (0.79 [0.58]; 0.85 [0.67]) walking. Similarly, there were no significant difference in tibialis anterior RMS EMG (P = .458, g = 0.293) for passive (0.71 ± 0.66) compared with the powered prosthesis (0.60 [0.82]). For intact limb quadriceps EMG, no significant differences were found for the vastus lateralis (P = .052, g = 0.976) or rectus femoris (P = .050, g = 0.994) between prosthetic conditions, but the effect sizes were large. The RMS EMG mean (standard deviation) was 0.59 (0.56) and 0.53 (1.15) for the vastus lateralis and 0.85 (1.67) and 0.77 (3.47) for rectus femoris during passive and powered prosthesis walking, respectively. Finally, biceps femoris muscle activity was not statistically different (P = .080) between the passive (0.58 [0.58]) and powered (0.60 [1.02]) prosthesis conditions.
No significant differences were found in user’s preference (preference score of 5) for the 2 prostheses. Participants had no preference (P = .225, g = 0.475) between walking with their passive prosthesis and our prosthesis under myoelectric control (preference score of 6.5 [1.1]) when walking outdoors. Individual participant’s preferences are presented in Table 3. All but one participant favored the powered prosthesis and scored 6.3 or higher.
Participant Prosthetic Preference
| Participant number | Prosthetic preference score |
|---|---|
| 1 | 8.1 |
| 2 | 8.6 |
| 3 | 7.9 |
| 4 | 7.4 |
| 5 | 6.3 |
| 6 | 1.6 |
Note: A score below 5.0 is toward participant’s prescribed prosthesis, while a score above 5.0 is toward the prosthesis under myoelectric control. A score of 5.0 indicates no difference between prostheses.
Discussion
Contrary to our hypothesis, participants had similar cost of transport when walking outdoors with the portable bionic ankle under proportional myoelectric control when compared with walking with their own prescribed passive prosthesis. When analyzing only the uphill and downhill sections of the course, there were no significant differences in cost of transport between walking with the 2 prostheses. The powered prosthesis did add mass (0.68 [0.031] kg) to the participants’ leg compared with their own passive prosthesis, which likely contributed to the slight increase in metabolic cost of transport with the bionic prosthesis.41,42 From our past work on laboratory-based biomechanics measurements with the same participants and prostheses, there was a 24% increase in peak ankle power for walking with the bionic prosthesis compared with the passive prostheses.24 This suggests that the added mass may have countered the increased ankle push-off achieved with the bionic prosthesis.
Although other studies with powered ankle prostheses have shown reductions in metabolic costs, few studies evaluated powered prostheses in real-world environments outside of the laboratory. Previous studies measured metabolic cost or cost of transport for level ground or incline/decline walking on a treadmill9,43–45 or on a track,4 which does not simulate uneven/cracked sidewalks or other terrain obstacles present in real-world environments. Studies have biomechanically evaluated powered prosthesis use over loose rocks and activity of daily living obstacle courses,20,46 but did not include metabolic measures. Kim et al investigated participant mobility, preference, daily activity, and metabolic cost with the BiOM powered ankle compared with their prescribed prosthesis after 2 weeks of at home use and found no significant changes in metabolic cost on a treadmill, or, perception of mobility in real-world use.10 Their preference metrics often did not correlate with measures of metabolic cost, which was similar to our results where 5 of 6 participants preferred walking with myoelectric control but without reducing their cost of transport compared with their passive prosthesis. Removing participant 6’s preference score of 1.6 would increase the preference average to 7.7 out of 10, and preference toward the prosthesis under myoelectric control would have been statistically significant (P = .005). Additionally, in our study, multiple participants indicated they felt it was easier to walk uphill; however, the cost of transport data did not match their perception. This highlights that prosthetic preference is not directly driven by cost of transport, thus quantitative and qualitative measures should be used in tandem for prosthetic evaluation.
While walking with the powered prosthesis outdoors, we saw significant increases in muscle activity for several muscles compared with passive prosthesis walking. For the intact limb, there was a significant increase in semitendinosus muscle activity when walking with the powered compared with the passive prosthesis. Participants may have been trying to stabilize themselves with their intact limb when walking with the powered prosthesis. For the residual limb, the vastus lateralis and rectus femoris muscles were significantly more active in early and mid stance (0% to 40% of the gait cycle) walking with the powered compared to the passive prosthesis (Figure 4). Individuals with amputation typically compensate for loss of ankle muscles through increased residual hamstring and rectus femoris activity during early stance and preswing phases compared with healthy controls.47–50 Previous work by Kim et al51 evaluated muscle activity and metabolic costs of individuals with transtibial amputation walking with the BiOM powered prosthesis and found significantly increased intact limb gluteus medius and residual limb vastus medialis when walking with the BiOM compared with a passive device. Additionally, they found nonsignificant, but moderate, to strong, correlations for metabolic cost reduction walking with the BiOM with increased residual rectus femoris activity in the residual thigh muscles during terminal stance compared with a passive prosthesis. We primarily see increased quadriceps and hamstring residual limb muscle activity during early to midstance walking with the myoelectric prosthesis compared with participants’ passive device. However, we could draw similar conclusions to Kim et al that participants may increase residual limb rectus femoris activity during early/midstance as a strategy to stabilize themselves during weight acceptance and to stiffen the knee to utilize ankle power. For our participants, we saw significant increases in quadriceps muscle activity between powered and passive prostheses, which could be a mechanism to relax their residual gastrocnemius to correctly time the myoelectric controller plantar flexion. Previous studies walking with a myoelectric ankle prosthesis did not record thigh EMG to compare results. Additionally, the increased quadriceps and hamstring muscle activity could have been related to participants’ still familiarizing themselves with how to walk with our powered prosthesis. Studies suggest individuals with transtibial amputation needed over a week of prosthesis familiarization time to measure significant changes in metabolic cost and reduce variability in prosthetic ankle joint kinematics.6,52,53 Our participants were limited to several hours (3–6 h) of walking time over multiple sessions in the powered prosthesis before the outdoor walking evaluation, with no means for additional in home training. Although individuals with amputation have adapted to new prostheses within a couple hours, this seems to translate when the new prosthesis has a similar motion pattern to the users’ prescribed prosthesis.54 Myoelectric control can require a complete readjustment in muscle recruitment timing. The user directly modulates prosthetic dynamics, which, at first, can produce very different motion patterns than a user’s prescribed prosthesis. Developing a myoelectric prosthesis robust enough for home use would allow us to further understand adaptation to walking with myoelectric control.
There were limitations for this study. First, we controlled outdoor walking speed based on the slowest self-selected pace between the powered and passive prosthesis for each participant. Having the participants walk at the same pace for both prostheses masked any differences related to differences in preferred walking speed between the 2. We did not measure iliopsoas muscle activity, which might have indicated potential compensations for the added distal weight from the powered prosthesis. The limited range of motion and power capabilities of the Open Source prosthesis were not ideal. They were both less than intact human values. In the future, a bionic lower limb prosthesis that can achieve full range of motion and power output would provide a better replacement for amputees. Lastly, the respiratory exchange ratio measures using breath by breath mode with the COSMED sometimes drifted above 1.0 during the data collections walking outside. The respiratory exchange ratio values greater than 1.0 were not long in duration, and there was no systematic difference between the 2 prosthesis testing conditions and may also be due to more variability using breath by breath measures.
Overall, there were no significant differences in the cost of transport for outdoor walking between a prosthesis under continuous myoelectric control and a passive prosthesis. There was significantly increased residual quadriceps muscle activity primarily during early stance, which could be attributed to increased residual limb stabilization with the powered prosthesis. Although the cost of transport increased when walking outdoors with a prosthesis under myoelectric control compared with a passive prosthesis, 5 out of 6 participants preferred walking with the myoelectric prosthesis. Therefore, future studies should continue to measure perception and other subjective and objective measures to evaluate potential advancements in prostheses technology.
Acknowledgments
The authors would like to thank Cara Negri, BSME, FAAOP for her time and contribution to participant prosthetic alignments. Additionally, the authors would like to thank the participants for their time and effort for contributing to the study. We would also like to thank other members of the University of Florida Human Neuromechanics Laboratory for assisting with various aspects of the project. This research was supported by NIH T32 HD043730 and NSF BCS-1835317.
References
- 2.
Houdijk H, Pollmann E, Groenewold M, Wiggerts H, Polomski W. The energy cost for the step-to-step transition in amputee walking. Gait Posture. 2009;30(1):35–40. doi:
- 3.↑
Hsu MJ, Nielsen DH, Lin-Chan SJ, Shurr D. The effects of prosthetic foot design on physiologic measurements, self-selected walking velocity, and physical activity in people with transtibial amputation. Arch Phys Med Rehabil. 2006;87(1):123–129. doi:
- 4.↑
Au SK, Weber J, Herr H. Powered ankle-foot prosthesis improves walking metabolic economy. IEEE Trans Robot. 2009;25(1):51–66. doi:
- 5.↑
Herr HM, Grabowski AM. Bionic ankle-foot prosthesis normalizes walking gait for persons with leg amputation. Proc R Soc B Biol Sci. 2012;279(1728):457–464. doi:
- 6.↑
Grabowski AM, Rifkin J, Kram R. K3 promoter™ prosthetic foot reduces the metabolic cost of walking for unilateral transtibial amputees. J Prosthetics Orthot. 2010;22(2):113–120. doi:
- 7.↑
Ingraham KA, Choi H, Gardinier ES, Remy CD, Gates DH. Choosing appropriate prosthetic ankle work to reduce the metabolic cost of individuals with transtibial amputation. Sci Rep. 2018;8(1):569. doi:
- 8.↑
Esposito ER, Whitehead JMA, Wilken JM. Step-to-step transition work during level and inclined walking using passive and powered ankle-foot prostheses. Prosthet Orthot Int. 2016;40(3):311–319. doi:
- 9.↑
Gardinier ES, Kelly BM, Wensman J, Gates DH. A controlled clinical trial of a clinically-tuned powered ankle prosthesis in people with transtibial amputation. Clin Rehabil. 2018;32(3):319–329. doi:
- 10.↑
Kim J, Wensman J, Colabianchi N, Gates DH. The influence of powered prostheses on user perspectives, metabolics, and activity: a randomized crossover trial. J Neuroeng Rehabil. 2021;18(1):842. doi:
- 11.↑
Fey NP, Simon AM, Young AJ, Hargrove LJ. Controlling knee swing initiation and ankle plantarflexion with an active prosthesis on level and inclined surfaces at variable walking speeds. IEEE J Transl Eng Health Med. 2014;2:228. doi:
- 12.
Gehlhar R, Tucker M, Young AJ, Ames AD. A review of current state-of-the-art control methods for lower-limb powered prostheses. Annu Rev Control. 2023;55:142–164. doi:
- 13.↑
Bhakta K, Maldonado-Contreras J, Camargo J, et al. Multi-context, user-independent, real-time intent recognition for powered lower-limb prostheses. 2023. https://hdl.handle.net/1853/70300. Accessed September 27, 2023.
- 14.↑
Fleming A, Stafford N, Huang S, Hu X, Ferris DP, Huang HH. Myoelectric control of robotic lower limb prostheses: a review of electromyography interfaces, control paradigms, challenges and future directions. J Neural Eng. 2021;18(4):41004. doi:
- 15.↑
Chen B, Wang Q, Wang L. Adaptive slope walking with a robotic transtibial prosthesis based on volitional EMG control. IEEE/ASME Trans Mechatron. 2015;20(5):2146–2157. doi:
- 16.
Creveling S, Cowan M, Sullivan LM, Gabert L, Lenzi T. Volitional EMG control enables stair climbing with a robotic powered knee prosthesis. 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). 2023:2152–2157.
- 17.↑
Kannape OA, Herr HM. Volitional control of ankle plantar flexion in a powered transtibial prosthesis during stair-ambulation. 2014 36th Annu Int Conf IEEE Eng Med Biol Soc EMBC. 2014;10:1662–1665. doi:
- 18.↑
Bhakta K, Camargo J, Kunapuli P, Childers L, Young A. Impedance control strategies for enhancing sloped and level walking capabilities for individuals with transfemoral amputation using a powered multi-joint prosthesis. Mil Med. 2020;185(suppl 1):490–499. doi:
- 19.
Sullivan LM, Creveling S, Cowan M, Gabert L, Lenzi T. Powered knee and ankle prosthesis control for adaptive ambulation at variable speeds, inclines, and uneven terrains. Rep U S. 2023;10:2128–2133. doi:
- 20.↑
Meier MR, Hansen AH, Gard SA, McFadyen AK. Obstacle course: users’ maneuverability and movement efficiency when using Otto Bock C-Leg, Otto Bock 3R60, and CaTech SNS prosthetic knee joints. J Rehabil Res Dev. 2012;49(4):583–596. doi:
- 21.↑
Huang S, Wensman JP, Ferris DP. Locomotor adaptation by transtibial amputees walking with an experimental powered prosthesis under continuous myoelectric control. IEEE Trans Neural Syst Rehabil Eng. 2016;24(5):573–581. doi:
- 22.↑
Azocar AF, Mooney LM, Duval JF, Simon AM, Hargrove LJ, Rouse EJ. Design and clinical implementation of an open-source bionic leg. Nat Biomed Eng. 2020;4(10):941–953. doi:
- 23.↑
Azocar AF, Mooney LM, Hargrove LJ, Rouse EJ. Design and characterization of an open-source robotic leg prosthesis. Proceedings IEEE RAS EMBS International Conference Biomedical Robot Biomechatronics. 2018:111–118. doi:
- 24.↑
Stafford N, Gonzalez EB, Ferris D. Walking biomechanics of individuals with transtibial amputations using a prescribed prosthesis and a portable bionic prosthesis under myoelectric control. Authorea Prepr. 2024;10:799. doi:
- 25.↑
Fleming A, Huang S, Buxton E, Hodges F, Huang HH. Direct continuous electromyographic control of a powered prosthetic ankle for improved postural control after guided physical training: a case study. Wearable Technol. 2021;2:3. doi:
- 26.
Koller JR, David Remy C, Ferris DP. Comparing neural control and mechanically intrinsic control of powered ankle exoskeletons. IEEE Int Conf Rehabil Robot. 2017:20;294–299. doi:
- 27.↑
Lenzi T, De Rossi SMM, Vitiello N, Carrozza MC. Intention-based EMG control for powered exoskeletons. IEEE Trans Biomed Eng. 2012;59(8):2180–2190. doi:
- 28.↑
Huang S, Wensman JP, Ferris DP. An experimental powered lower limb prosthesis using proportional myoelectric control. J Med Device. 2014;8(2):24501. doi:
- 29.
Fleming A, Huang S, Buxton E, Hodges F, Huang H. Direct continuous EMG control of a powered prosthetic ankle for improved postural control after guided physical training: a case study. Biorxiv. 2020;10:293373. doi:
- 30.
Fleming A, Huang S, Huang H. Proportional myoelectric control of a virtual inverted pendulum using residual antagonistic muscles: toward voluntary postural control. IEEE Trans Neural Syst Rehabil Eng. 2019;27(7):1473–1482. doi:
- 31.↑
Fleming A, Liu W, Huang HH. Neural prosthesis control restores near-normative neuromechanics in standing postural control. Sci Robot. 2023;8(83):eadf5758. doi:
- 32.↑
Hermens HJ, Freriks B, Merletti R, et al. European recommendations for surface electromyography. Roessingh Res Dev. 1999;8(2):13–54.
- 33.↑
Huang S, Ferris DP. Muscle activation patterns during walking from transtibial amputees recorded within the residual limb-prosthetic interface. J Neuroeng Rehabil. 2012;9(1):55. doi:
- 34.↑
Selinger JC, Donelan JM. Estimating instantaneous energetic cost during non-steady-state gait. J Appl Physiol. 2014;117(11):1406–1415. doi:
- 35.↑
Jasiewicz JM, Allum JHJ, Middleton JW, et al. Gait event detection using linear accelerometers or angular velocity transducers in able-bodied and spinal-cord injured individuals. Gait Posture. 2006;24(4):502–509. doi:
- 36.↑
Bastas G, Fleck JJ, Peters RA, Zelik KE. IMU-based gait analysis in lower limb prosthesis users: Comparison of step demarcation algorithms. Gait Posture. 2018;64:30–37. doi:
- 37.↑
Brockway JM. Derivation of formulae used to calculate energy expenditure in man. Hum Nutr Clin Nutr. 1987;41(6):463–471. https://europepmc.org/article/med/3429265. Accessed June 26, 2022.
- 38.↑
Hybart R, Villancio-Wolter KS, Ferris DP. Metabolic cost of walking with electromechanical ankle exoskeletons under proportional myoelectric control on a treadmill and outdoors. PeerJ. 2023;11:e15775. doi:
- 39.↑
MacLean MK, Ferris DP. Energetics of walking with a robotic knee exoskeleton. J Appl Biomech. 2019;35(5):320–326. doi:
- 40.↑
Fritz CO, Morris PE, Richler JJ. Effect size estimates: current use, calculations, and interpretation. J Exp Psychol Gen. 2012;141(1):2–18. doi:
- 41.↑
Browning RC, Modica JR, Kram R, Goswami A. The effects of adding mass to the legs on the energetics and biomechanics of walking. Med Sci Sports Exerc. 2007;39(3):515–525. doi:
- 42.↑
Smith JD, Martin PE. Effects of prosthetic mass distribution on metabolic costs and walking symmetry. J Appl Biomech. 2013;29(3):317–328. doi:
- 43.↑
Montgomery JR, Grabowski AM. Use of a powered ankle–foot prosthesis reduces the metabolic cost of uphill walking and improves leg work symmetry in people with transtibial amputations. J R Soc Interface. 2018;15(145):442. doi:
- 44.
Darter BJ, Wilken JM. Energetic consequences of using a prosthesis with adaptive ankle motion during slope walking in persons with a transtibial amputation. Prosthet Orthot Int. 2013;38(1):84. doi:
- 45.↑
Hafner BJ, Halsne EG, Morgan SJ, Morgenroth DC, Humbert AT. Effects of prosthetic feet on metabolic energy expenditure in people with transtibial amputation: a systematic review and meta-analysis. PM&R. 2021;10:693. doi:
- 46.↑
Gates DH, Aldridge JM, Wilken JM. Kinematic comparison of walking on uneven ground using powered and unpowered prostheses. Clin Biomech. 2013;28(4):467–472. doi:
- 47.↑
Fey NP, Silverman AK, Neptune RR. The influence of increasing steady-state walking speed on muscle activity in below-knee amputees. J Electromyogr Kinesiol. 2010;20(1):155–161. doi:
- 48.
Winter DA, Sienko SE. Biomechanics of below-knee amputee gait. J Biomech. 1988;21(5):361–367. doi:
- 49.
Isakov E, Burger H, Krajnik J, Gregoric M, Marincek C. Knee muscle activity during ambulation of trans-tibial amputees. J Rehabil Med. 2001;33(5):196–199.
- 50.↑
Powers CM, Rao S, Perry J. Knee kinetics in trans-tibial amputee gait. Gait Posture. 1998;8(1):16. doi:
- 51.↑
Kim J, Gardinier ES, Vempala V, Gates DH. The effect of powered ankle prostheses on muscle activity during walking. J Biomech. 2021;124:110573. doi:
- 52.↑
Wurdeman SR, Myers SA, Jacobsen AL, Stergiou N. Adaptation and prosthesis effects on stride-to-stride fluctuations in amputee gait. PLoS One. 2014;9(6):e100125. doi:
- 53.↑
Delussu AS, Brunelli S, Paradisi F, et al. Assessment of the effects of carbon fiber and bionic foot during overground and treadmill walking in transtibial amputees. Gait Posture. 2013;38(4):876–882. doi:
- 54.↑
Schmalz T, Bellmann M, Proebsting E, Blumentritt S. Effects of adaptation to a functionally new prosthetic lower-limb component: results of biomechanical tests immediately after fitting and after 3 months of use. J Prosthetics Orthot. 2014;26(3):134–143. doi:
