Development and Preliminary Evaluation of a Bimodal Foot Prosthesis for Walking and Running Open Access

People often alternate between bouts of walking and running, for instance, when adults participate in recreational activities or when children play [1,2]. However, individuals with lower limb prostheses have reported that transitioning between walking and running activities can be challenging because existing prosthetic feet are not well-suited for both activities [3,4]. Individuals with lower limb prostheses have also reported that they may forego taking part in play or recreational activities to avoid using a foot not well-suited for the activity [4,5]. Thus, limitations of prosthetic feet can deter participation in physical activity, which may negatively impact social or physical health [4–7].

Some prosthesis users wear a daily-use prosthesis (DUP) for walking and standing, but then wear a separate, specialized running-specific prosthesis (RSP) for running. Switching between the two prostheses is often impractical or inconvenient, especially when switching activities frequently or unexpectedly [4]. Switching prostheses can be time consuming and can require specialized tools (e.g., hex keys). Switching prostheses also requires users to have their other prosthesis available, which means they either need to anticipate each activity they will perform each day or carry their second prosthesis (sometimes with an associated socket) throughout the day. Situations inevitably arise when an individual wearing a DUP wants or needs to run (e.g., an unplanned game of frisbee). Similarly, an individual wearing an RSP may need to stand or walk unexpectedly (e.g., when they are out for a run but stop to talk to an acquaintance). Using a DUP for running or using an RSP for walking (or standing) can result in dissatisfied users and decreased performance in the activity [3,4].

These performance detriments are due to the differences in the designs of the DUP versus RSP. RSPs are generally made of long, curved carbon fiber keels that are stiffer than the DUPs and are designed for the higher forces associated with running (more than twice the forces produced during walking [8]). The RSP's specialized design comes at the cost of task versatility. RSPs typically omit a heel component so users can comfortably contact the ground with the midfoot of the prosthesis during running. However, the lack of heel makes daily activities like walking and standing less stable and less comfortable [4,9]. Similarly, the DUP designs are not well-suited for running as they are generally less stiff and yield lower energy return [10]. Also, the heel of DUP can interfere with a user's midfoot ground contact during running. Some companies offer crossover prostheses that are designed for walking, running, and more, but these prosthetics provide a single dynamic response profile for all activities and thus do not fully meet the varied needs of all users or activities [11].

We posit that prosthesis users might benefit from having one foot that can easily and quickly switch between running and walking modes to address the time-consuming impracticality of switching between a DUP and RSP. The aim of this work was to create and evaluate a bimodal prosthetic foot prototype for walking and running. We employed a user-centric design approach to understand user needs, collect user feedback, and iteratively develop a prototype of the bimodal prosthesis. We assessed the prototype via benchtop testing, gait analysis on three prosthesis users, and user surveys. We evaluated the bimodal prosthesis prototype based on four primary criteria: (i) its ability to switch between walking and running modes faster than switching between DUP and RSP feet, (ii) its ability to provide a heel component in walking mode but remove it in running mode, (iii) its ability to generate higher axial stiffness in running mode, and (iv) user satisfaction with the bimodal prosthesis compared to standing and walking in their RSP, and compared to running in their DUP. Additional design requirements and constraints are detailed in the Methods. Secondarily, we explored how the different prostheses affected gait biomechanics, specifically center of mass and prosthesis power, when walking and running with the bimodal prosthesis, a DUP, and an RSP. We also characterized the stiffnesses of participants' DUPs and RSPs to build upon the limited data available in the literature.

We compiled user needs and engineering design requirements, as summarized below.

We established user needs based on a review of the academic literature [12,13], our previous interactions with dozens of prosthesis users and their caregivers [3], and a series of conversations with one prosthetist and with five prosthesis users who own both DUPs and RSPs (two unilateral transtibial, two bilateral transtibial, one bilateral transtibial and transfemoral). This approach enabled us to prioritize a design focused on scientifically documented user needs as well as the insights of end-users who would use a bimodal prosthesis. We do acknowledge that the feedback we received, as well as the user needs identified in prior literature, may not be comprehensive because they are often limited by the types of people who choose to volunteer to participate in research studies or to discuss their needs with researchers.

Users desired a broad footprint (area of contact with the ground) to improve stability while doing daily tasks like standing and walking [12,14]. Users noted that standing and walking in the RSPs throughout the day leads to discomfort and hip height misalignment [4] since RSPs are typically 2.8–4.5% taller than DUPs [15]. Therefore, having a prosthesis that was the same height as a DUP would help avoid introducing hip height misalignment. Users preferred DUPs that were less stiff than their RSPs and attributed the lower stiffness to lower effort during walking.

Users preferred RSPs that were stiffer and felt “springier” than their DUPs. They attributed the higher stiffness to improved ground push-off during running. Prosthesis users preferred RSPs that could provide high energy return for running [13,16,17]. Finally, users preferred RSPs without a heel component because it allows for greater heel clearance, leading to easier midfoot striking and a more natural rollover for running [9,16].

Some prosthesis users have comorbidities that lower their strength and dexterity [6]. Switching between the device modes, therefore, needs to be simple and to require minimal gestures to make the device usable by people of varying abilities. Switching modes also needs to be fast and require no tools. Users generally preferred not to have extra components that needed to be carried separately.

We derived engineering requirements from the user needs described above and created target values based on the scientific literature.

During standing, we required that the device have a broad footprint that mimics the width of an RSP (about 8.5 cm) and the length of a DUP (about 24 cm) for a minimum footprint of 204 cm². The widths of DUPs are smaller than the user's actual footprint, as they are typically worn with a cosmetic foot shell and shoe that increases the overall footprint width. Instead of the width of the DUP, the width of an RSP is a more appropriate target since, like the bimodal prosthesis, RSPs are commonly worn without a shell or shoe, representing a sufficient width for the user's footprint. A broad footprint gives users a base of support comparable to a shod DUP and offers enhanced stability during standing and walking [14].

We required the device to have the same build height as a DUP during walking mode to avoid limb length discrepancies and their associated musculoskeletal health risks [18].

We required the device to be less stiff when in walking mode than in running mode to provide users with loading and push-off dynamics closer to that of a typical DUP.

We required the device to be stiffer in running mode than in walking mode so that the user experiences similar loading and push-off dynamics compared to a typical RSP. Previous research suggests 20 kN/m and 26 kN/m are exemplary DUP and RSP stiffnesses, respectively, for an 80 kg prosthesis user [19,20].

We required the device to have more heel clearance in running mode compared to walking mode to enable the user to make ground contact with their fore-/midfoot. We estimated that a 5 cm increase in heel height would mimic the rollover shape of the Obsidian (Fillauer, Chattanooga, TN) RSP and allow users to make ground contact with their fore-/midfoot in running.

We required that the device mode switching would not need extra tools and that it could be performed quickly (less than a minute).

Device mass is important to consider as excessive mass would increase the device's inertial effects and the effort associated with walking or running [21]. Commercial DUPs typically have a mass of approximately 1 kg and some commercial RSPs may be even lighter. We selected a target of 2 kg for the device as a compromise between functionality and mass for an early-stage prototype, with the expectation that if this concept is promising, the device could be optimized for minimal mass in future work.

We designed and fabricated a passive bimodal prosthesis with two main capabilities: (i) the ability to switch between different walking and running stiffnesses and (ii) the ability to provide a heel component for walking and standing, but not for running. We fabricated a prototype to satisfy the user needs and technical requirements above by modifying a commercial DUP (direct mount AllPro, Fillauer, Chattanooga, TN). The modified prosthesis was a size 26 cm, 254 mm (10″) tall prosthesis intended for users of 73 to 90 kg.

We fabricated a parallel spring strut and locking joint (Fig. 1) that allows the user to switch between higher stiffness for running and lower stiffness for walking. While unlocked (Fig. 1), the joint flexes under the user's weight without compressing the parallel spring strut. Therefore, the device's overall stiffness during walking is just the intrinsic stiffness of the original DUP keel. While locked (Fig. 2), the joint does not flex under the user's weight, compressing the parallel spring strut and increasing the overall stiffness of the device in running mode. Users may change springs inside the parallel spring strut to achieve their desired stiffness for running.

We replaced the commercial DUP footplate with a custom-built, rounded aluminum toeplate that mimicked the profile of an RSP (Obsidian Running Blade, Fillauer, Chattanooga, TN). We attached a collapsible heel via a hinge joint to the toeplate. A heel spring strut then slides along the collapsible heel via channels (Fig. 1). The other end of the heel spring strut is hinged to the back of the prosthesis keel. A bolt latch locks the heel spring strut in place on either the anterior or posterior end of the collapsible heel. In walking mode, the collapsible heel is parallel to the ground and the heel spring strut is locked to the posterior end of the heel, increasing the device's footprint and assisting in weight acceptance following initial ground contact (Fig. 1). In running mode, the collapsible heel folds up against the back of the prosthesis keel and the heel spring strut is locked in the anterior end of the heel, increasing heel clearance during running (Fig. 2). We also lined the bottom of the aluminum toeplate and the collapsible heel with a layer of dense foam (shoe sole material) to increase traction against the ground.

This design allows the user to switch between a walking mode and a running mode. The user can manually unlock the joint above the parallel spring strut and extend the collapsible heel for walking mode (Fig. 1). The user can then lock the joint above the parallel spring strut and collapse the heel for running mode (Fig. 2). An exploded view of the device (Fig. 3) illustrates how the commercial prosthesis was modified to include the parallel spring strut, locking joint, and heel spring strut.

We completed benchtop measurements to assess if the prototype satisfied a subset of the user needs and engineering requirements. We then conducted a case series study (N = 3) where lower limb prosthesis users walked and ran in their DUPs, RSPs, and the bimodal prosthesis prototype to assess the remaining requirements, particularly related to functionality and user satisfaction.

We evaluated the bimodal prosthesis prototype based on the four primary criteria outlined in the Introduction.

Benchtop measurements included finding the prototype's mass and dimensions (build height, footprint, and heel clearance). The device's mass was measured using a scale. The length of the footprint was measured as the distance between the posterior end of the heel in walking mode and the most anterior point of the toe plate that was in contact with the ground (not including the curved portion). The width of the footprint was measured as the distance between the medial and lateral edges of the prosthesis. We calculated the area of the footprint by multiplying the measured length by the width and compared it to the target footprint from the design requirements (8.5 cm × 24 cm). We measured the heel clearance as the height of the posterior end of the prosthesis heel in running mode. We compared the prototype heel clearance to the target heel clearance from the design requirements (5 cm).

We completed testing on three participants with unilateral transtibial limb loss. Each participant attended two sessions: one familiarization session, and one data collection session at Vanderbilt University's Center for Rehabilitation Engineering and Assistive Technology. The protocol was approved by the Vanderbilt University Institutional Review Board (#150271) and all participants gave their written informed consent.

Participants wore their prescribed socket and completed each task in all three prostheses (DUP, RSP, bimodal) while wearing an upper body safety harness (attached to the ceiling via a slightly slack cable). Participant demographics for the resulting group are provided in Table 1.

Participants familiarized themselves with the bimodal prosthesis while walking and running on the treadmill at 1.1, 2.0, and 2.5 m/s for roughly 60 min during the first session (about 15–20 min of walking, 5–10 min of running, plus breaks and alignment adjustments per the participant's discretion). Participants also familiarized themselves with performing the switch between the bimodal prosthesis walking and running modes (about three times with each mode switch).

For both familiarization and data collection, we affixed ankle weights to the participants' prescribed DUP and RSP so that they were equal to the mass of the bimodal prosthesis (within ± 50 g). We set the walking speed to 1.1 m/s and the running speeds to 2.0 and 2.5 m/s, similar to speeds tested in previous DUP and RSP walking and running studies [22–24].

We conducted all data collection trials on a split-belt force-instrumented treadmill (Bertec, Columbus, OH), with a sampling frequency of 1000 Hz. We affixed retroreflective markers to the prostheses and the participant's lower body and their positions were recorded with a Vicon motion capture system (200 Hz; Vicon, Oxford, UK). Six markers were applied to the pelvis, four on each thigh, two on each knee, four on each shank (markers were adhered to the socket on the prosthesis-side). On the nonprosthesis-side, two markers were applied to the malleoli, two on the metatarsal heads, and three on the calcaneus. Prosthesis-side markers are described in the Stiffness Section. The marker set is shown in Fig. 4. We filtered ground reaction force (GRF) and synchronized motion capture data with zero-phase, 4th order, low-pass Butterworth filters with cutoff frequencies of 15 and 8 Hz, respectively.

Participants stood for 15 s in a static position (arms crossed, feet shoulder width apart), walked for 90 s, and ran 45 s at both running speeds (except for P3 who could not run at 2.5 m/s) with each of the prostheses. We analyzed only the final 60 s of walking data and the final 30 s of running data to ensure processing of constant speed locomotion data.

We asked participants to run with three different compression springs (stiffness: 6.1, 19.3, and 29.8 kN/m) for the bimodal parallel spring during the first session. The 6.1 kN/m spring was provided to allow users to approximate the difference between some exemplary DUP and RSP stiffnesses (20 kN/m and 26 kN/m) for an 80 kg user [19,20]. The 19.3 and 29.8 kN/m springs were provided for users that preferred even stiffer RSPs. The specific values of the two stiffer springs were informed by pilot testing and based on the availability of commercial-off-the-shelf springs that fit into the Parallel Spring Strut. We then allowed them to choose their preferred stiffness for running during the data collection session. P1 chose a 29.8 kN/m spring, and P2 and P3 chose a 6.1 kN/m spring.

We characterized the forefoot stiffness of each prosthesis between the proximal end of the prosthesis and the toe region during each task. We measured the prosthesis displacement by recording the distance between two markers: one affixed below the direct mount pyramid adapter and one located a few centimeters posterior to the distal end of the keel (Fig. 5). We defined the displacement axis as the line along the markers in the sagittal plane.

We calculated the prosthesis resting length 10 frames before ground contact for each step while the foot was unloaded. We calculated stiffness at the instant where the center of pressure on the treadmill belt was closest to the displacement axis. We calculated the force along the displacement axis as the dot product of the GRF and displacement axis unit vector. We divided the force along the displacement axis by the difference between the resting length and the distance between the markers at that instant to calculate the device's forefoot stiffness. This calculation assumes a linear relationship between forefoot force and displacement; linear characterizations have been shown to be strongly predictive of the curvilinear force–displacement profiles of RSPs [19]. We compared the walking mode stiffness to the running mode stiffness to assess the prototype's ability to generate higher axial stiffness in running mode.

We averaged each user's bimodal prosthesis running mode stiffnesses across the 2.0 and 2.5 m/s running tasks to compare against the walking mode stiffness from their 1.1 m/s walking trials. We averaged each user's DUP stiffness across all three tasks (1.1, 2.0, and 2.5 m/s). We also averaged each user's RSP stiffnesses across all three tasks for comparison. We characterized the stiffnesses of participants' DUPs and RSPs to build upon the limited existing data in the literature.

We asked participants to complete a survey after each task. Survey questions utilized a seven-point Likert scale, the use of which has been previously established for prosthesis users [25]. A rating of 1 on the Likert scale indicated that the user perceived the device being assessed as “very bad” in that attribute, a 4 indicated the device was “neutral,” and a 7 indicated the device was “very good.” Users rated each foot in terms of comfort, energy return, stability, hip height, and overall satisfaction. We did not ask users to rate the energy return of the feet for standing.

We also asked participants to rate the ease of switching between their DUP and RSP, and the ease of switching between the bimodal prosthesis modes. We timed users while switching modes and compared these times to assess the prototype's ability to switch between running and walking modes faster than switching between DUP and RSP feet.

We subtracted the RSP survey ratings from the bimodal prosthesis walking mode ratings during standing and walking tasks. We subtracted the DUP ratings from the bimodal prosthesis running mode ratings during running tasks. Hence, a positive value indicates that the user thought the bimodal prosthesis was better for that attribute than a prosthesis that was not designed for that task. We computed these differences to assess user satisfaction with the bimodal prosthesis compared to standing and walking in their RSP, and compared to running in their DUP as shown in the Results. Comparison by subtraction of Likert scale ratings has also been previously established for prosthesis users [25].

For completeness, we also subtracted DUP survey ratings from the bimodal prosthesis walking mode ratings during standing and walking tasks. Finally, we subtracted RSP survey ratings from the bimodal prosthesis running mode ratings during running tasks. All ratings and differences can be found in Appendix  A.

We computed the participant's center of mass power (individual limbs method [26]) and the prosthesis power (unified deformable method [27]) for each trial. We selected the center of mass power to use as a representative metric for gait mechanics as it indicates the whole-body effects of each prosthesis and serves as a single signal to compare across devices. We computed prosthesis power as a secondary metric to characterize the mechanical response of each prosthesis during the locomotion tasks. We cropped the center of mass power to a stride, time normalized to percent stride, and averaged across strides. A stride was defined as starting with prosthesis ground contact and ending just before the subsequent prosthesis ground contact.

Participant preferences varied between switching bimodal modes or switching devices (Fig. 6). In terms of perceived ease of switching, one participant rated the bimodal prosthesis modes easier to switch than switching their DUP and RSP, one gave the two methods the same rating, and one rated the bimodal prosthesis modes harder than switching their DUP and RSP.

Participants switched between bimodal prosthesis modes faster than they switched between their DUP and RSP (i.e., doffing one and donning the other). On average, users switched 29.5 s (59%) faster: 21.3 ± 12.0 s (10–40 s) to switch between bimodal prosthesis modes versus 50.8 ± 34.4 s (20–78 s) to switch between their DUP and RSP. The two prosthesis users who took the longest to switch between DUP and RSP (52–78 s) had to use a hex key tool to detach and reattach their prostheses. However, one participant (P1) was able to switch his DUP and RSP in 20 to 28 s because he carried an extra foot with its own socket and therefore could doff one socket and then don the other without tools (though at the cost of carrying an extra prosthetic socket).

The bimodal prosthesis footprint in walking mode was 234.4 cm² and achieved the target footprint of at least 204.4 cm² (14.9% greater). The heel clearance of the bimodal prosthesis in running mode was 6.7 cm and achieved the target of at least 5 cm.

We found that the bimodal prosthesis height in walking mode was the same as the DUP height while in a shoe and therefore did not introduce limb length discrepancies. Switching modes could be accomplished using only one finger or thumb and took only five gestures (twist and slide the lock sheath; unlock, slide, and lock the heel spring strut). The bimodal prosthesis has a total build mass of 1.8 kg which achieved our target range of less than 2 kg for the initial prototype.

The bimodal prosthesis was stiffer in running mode than in walking mode. The stiffness increase ranged from 9.8 to 36.2 kN/m (23–84%; Fig. 7) across the three participants. The magnitude of increase was dependent on which parallel spring stiffness each participant selected based on their preference (e.g., P1 selected a much stiffer spring than P2 and P3, see Methods Section for details). The stiffness values measured at each running speed were similar (within ±5% at 2.0 versus 2.5 m/s, Fig. 7).

During standing, participants preferred the bimodal prosthesis in walking mode over their RSP. In terms of overall user satisfaction, all participants rated the bimodal prosthesis walking mode to be higher than the RSP, by 2.3 points on average (6.3 versus 4.0, i.e., moderately to very good versus neutral satisfaction). We observed similar trends for all other survey attributes (comfort, stability, hip height), excluding energy return which is not applicable to standing. On average, users rated the bimodal prosthesis walking mode to be 2.3–3.7 points higher in each attribute relative to the RSP (Fig. 8). In contrast, the users rated the bimodal prosthesis walking mode to be similar to their DUP (e.g., 6.3 versus 6.7 in terms of overall satisfaction). Complete participant-specific survey results for bimodal prosthesis versus RSP are reported in Table 2 and for bimodal versus DUP are reported in Table 6.

During walking, participants preferred the bimodal prosthesis in walking mode over their RSP. In terms of overall user satisfaction, all participants rated the bimodal prosthesis walking mode to be higher than the RSP, by 2.3 points on average (5.6 versus 3.3, i.e., slightly to moderately good versus slightly bad to neutral). We observed similar trends for all other survey attributes (comfort, stability, energy return, hip height). On average, users rated the bimodal prosthesis walking mode to be 1.7–2.3 points higher in each attribute relative to the RSP (Fig. 8). The users rated the bimodal prosthesis walking mode to be more similar to but slightly lower than their DUP (e.g., 5.7 versus 6.7 in terms of overall satisfaction). Complete participant-specific survey results for bimodal prosthesis versus RSP are reported in Table 3 and for and for bimodal versus DUP are reported in Table 7.

During running, participants preferred the bimodal prosthesis in running mode over their DUP. In terms of overall user satisfaction at 2.0 m/s, all participants rated the bimodal prosthesis in running mode to be higher than the DUP. At 2.5 m/s, one participant rated the bimodal prosthesis running mode to be higher than the DUP, and the second participant gave these devices the same rating. The third participant was unable to run at 2.5 m/s. On average, participants rated overall satisfaction with the bimodal prosthesis in running mode to be 1.7 points higher than the DUP at 2.0 m/s (6.0 versus 4.3, i.e., moderately good versus neutral to slightly good), and 0.5 points higher at 2.5 m/s (5.5 versus 5.0, i.e., slightly to moderately good versus slightly good; Fig. 8). We observed similar trends for all other survey attributes (comfort, stability, energy return, and hip height). On average, users rated the bimodal prosthesis running mode to be 0.3-1.7 points higher in each attribute relative to the DUP during 2.0 m/s running and to be 0.5–1.5 points higher during 2.5 m/s running. Users rated the bimodal prosthesis running mode to be similar to their RSP at 2.0 m/s (e.g., 6.0 versus 6.3 in terms of overall satisfaction, N = 3), but slightly lower than their RSP at 2.5 m/s (e.g., 5.5 versus 7 in terms of overall satisfaction, N = 2). Participant-specific survey results for bimodal versus DUP are reported in Tables 4 and 5 and for bimodal versus RSP are reported in Tables 8 and 9.

Biomechanical power and energy results were participant-specific. This is evident in Fig. 9, which shows intersubject variability in the center of mass power between the three participants (P1–P3) while running at 2.0 m/s. For example, for P1 the peak negative power magnitude when running on the bimodal prosthesis appears to better match the RSP condition, but for P3 the bimodal prosthesis appears to better match the DUP condition during this same phase of gait. For all walking and running conditions, we observed variable trends in power magnitudes and timing (of positive and negative power phases and peak power). There were no simple or consistent summary metrics to report. Instead, the participant-specific results for center of mass power and prosthesis power are provided in Appendices  B and  C (Figs. 11–13 in Appendix  B, Figs. 14–16 in Appendix  C).

RSP stiffness differed from DUP stiffness for each participant. The RSP was stiffer than the DUP for two of the three participants, and less stiff than the DUP for the remaining participant. The changes in stiffness from DUPs to RSPs ranged from −2.6 kN/m to +19.4 kN/m (−10% to +57%; Fig. 10) across P1, P2, and P3.

We successfully built and demonstrated a passive bimodal foot prosthesis for walking and running. Users could quickly switch the prosthesis prototype between walking and running modes (generally in less than 30 s). In walking mode (which was also used for standing), a heel component extended from the posterior end of the foot and the prosthesis had a nominal axial stiffness based on the carbon fiber foot properties. In running mode, the heel component collapsed out of the way and the user engaged a parallel spring strut that increased the axial stiffness of the prosthesis (by 23–84% relative to walking mode). During walking and standing, users were more satisfied with the bimodal prosthesis than their RSP and rated their satisfaction with the bimodal prosthesis to be similar to or slightly less than their satisfaction with their DUP. Likewise, during running, users were more satisfied with the bimodal prosthesis than their DUP and rated their satisfaction with bimodal prosthesis to be similar to or slightly less than their satisfaction with their RSP. These encouraging findings suggest that this bimodal prosthesis concept should be further developed and tested to better understand functional impacts, usability, and user acceptance over longer time durations and for more activities of daily living.

This bimodal prosthesis prototype is the first unpowered prosthesis with self-contained, modifiable foot shape configurations and stiffnesses to support both walking and running. There is previous academic research on modifiable stiffness prostheses, but the focus has mostly been on walking with passive, quasi-passive, semipowered [28–31], or powered prostheses [32,33]. One previous study investigated a prosthesis for walking and running, but it was powered and required an external power source [9]. Unlike powered prostheses, the passive prosthesis presented in this paper does not require motors or recharging. Prosthetics manufacturers have sometimes advertised certain prostheses as crossover or all-rounder feet with a single stiffness that is good for both walking and running; however, our conversations with prosthesis users (prior to this study) and study results (Figs. 8 and 10) indicate that users prefer substantially different stiffnesses for these different activities. Prosthetics manufacturers have also developed prostheses with either swappable heels or with variable stiffnesses, but not both. Thus, these prostheses still present logistical challenges or do not fully meet user needs for walking and running. For instance, the discontinued Catapult (Ottobock, Duderstadt, Germany) is an RSP with an insertable/removable parallel spring that allows users to adjust the forefoot stiffness. However, this prosthesis lacks a heel to support walking or standing balance. Furthermore, when the spring was removed, the user needed to carry around this extra component in case they wanted to change the stiffness back later. Alternatively, the Challenger (Ottobock, Duderstadt, Germany) is a DUP with an insertable/removable heel block, but it does not have adjustable forefoot stiffness and users still must carry the heel blocks throughout the day to adjust foot properties to accommodate different activities.

We characterized the center of mass and prosthesis power for participants walking and running on each type of prosthesis, but the results are difficult to interpret or generalize. There is no single or distinct feature in the center of mass power curve that determines how well a prosthesis performed or how much it benefited the user during locomotion. Nevertheless, these power results are presented in the Appendix to provide preliminary benchmark data to help inform future experimental methods and biomechanical outcomes. For the purposes of this study, we relied on user surveys and prosthesis stiffness measurements to more directly interpret prosthesis performance relative to user needs and engineering requirements.

Each participant's DUP differed in stiffness relative to their RSP, but not always in the direction we expected. For two participants, their RSP was up to +19.4 kN/m (+57%) stiffer than their DUPs. However, for the third participant, their RSP was −2.6 kN/m (−10%) less stiff than their DUP, which was counterto what we expected from prior literature [19,20]. We learned from conversations with prosthetists that there are situations when a user desires a less stiff RSP. Although the bimodal prosthesis prototype presented here was not developed with this in mind, fortunately, it could still be applied to these situations since the heel component and parallel spring strut can be adjusted independently. In other words, a user could collapse the heel and unlock the parallel spring strut if it were desirable to have a running mode with no heel and a lower stiffness than the walking mode.

Key limitations of this study include a short acclimation period to the bimodal prosthesis, a limited selection of parallel spring stiffnesses, and a prototype mass greater than a typical DUP or RSP. First, an hour of familiarization before testing may not have been sufficient for some participants to fully acclimate to walking and running with the device, or to learn how to efficiently switch between walking and running mode. Previous research shows that significant changes in walking balance occur within three weeks of receiving a new prosthesis [34]. A separate study found that individuals still make significant improvements with novel motor tasks (like switching between modes) two to five days after first learning the task [35]. In fact, P1 and P2 noted that, with a week more of practice, they believed they would rate the ease of switching between modes to be 2 and 1 points higher than the recorded ratings, respectively. Thus, the initial user survey ratings and biomechanical data may change over time and might not reflect the long-term results for the bimodal prosthesis. Second, the bimodal prosthesis prototype was only able to fit a small selection of springs in the parallel strut assembly because of the size constraints with the current prototype iteration and use of off-the-shelf springs. A future prototype could be designed to be compatible with a larger array of springs. A larger array of parallel springs could allow experimenters to tune the stiffness of the bimodal prosthesis to more closely match that of a user's RSP. Third, this prototype was designed with its own outsole tread. It was not designed to allow the user to wear footwear, which was a concern that some users expressed while commenting on the prototype. Future modifications to the bimodal prosthesis design or to footwear may facilitate the use of different footwear with this device, but this requires further exploration. Finally, the mass of the bimodal prosthesis (1.8 kg) was greater than that of the DUPs and RSPs (about 1.2 and 0.7 kg, respectively). We added ankle weights to the users' DUPs and RSPs so that their masses were equal to that of the bimodal prosthesis, but the increased mass may have affected their gait biomechanics [36].

We developed and tested a bimodal prosthesis prototype that can quickly switch between configurations and stiffnesses to accommodate standing, walking, and running. Preliminary results were encouraging in terms of user satisfaction, stiffness magnitude changes between modes, and the ease and speed of mode-switching. Future work should further develop this bimodal concept and explore user satisfaction associated with longer-term usage, a lighter prototype, and a wider range of activities.

We gratefully acknowledge Aaron Fitzsimmons and Joshua Southards from Amputee Blade Runners for their insights on the design and prescription of DUPs and RSPs and for their help in participant recruitment.

• NSF Graduate Research Fellowship Program (GRFP) (Grant No. 1937963; Funder ID: 10.13039/100000082).

DUP =

daily-use prosthesis

RSP =

running-specific prosthesis

GRF =

ground reaction force

All ratings and differences for the Bimodal compared to users' DUPs and RSPs.

Center of mass power calculations for each participant for walking (1.1 m/s) and running (2.0 and 2.5 m/s).

Prosthesis power calculations for each participant for walking (1.1 m/s) and running (2.0 and 2.5 m/s).