Transforming the Anthropomorphic Passive Free-Flow Foot Prosthesis Into a Powered Foot Prosthesis With Intuitive Control and Sensation (Bionic FFF)

ABSTRACT

Introduction

Approximately 89% of all service members with amputations do not return to duty. Restoring intuitive neural control with somatosensory sensation is a key to improving the safety and efficacy of prosthetic locomotion. However, natural somatosensory feedback from lower-limb prostheses has not yet been incorporated into any commercial prostheses

Materials and Methods

We developed a neuroprosthesis with intuitive bidirectional control and somatosensation and evoking phase-dependent locomotor reflexes, we aspire to significantly improve the prosthetic rehabilitation and long-term functional outcomes of U.S. amputees. We implanted the skin and bone integrated pylon with peripheral neural interface pylon into the cat distal tibia, electromyographic electrodes into the residual gastrocnemius muscle, and nerve cuff electrodes on the distal tibial and sciatic nerves. Results. The bidirectional neural interface that was developed was integrated into the existing passive Free-Flow Foot and Ankle prosthesis, WillowWood, Mount Sterling, OH. The Free-Flow Foot was chosen because it had the highest Index of Anthropomorphicity among lower-limb prostheses and was the first anthropomorphic prosthesis brought to market. Conclusion. The cats walked on a treadmill with no cutaneous feedback from the foot in the control condition and with their residual distal tibial nerve stimulated during the stance phase of walking

INTRODUCTION

Purpose of the Study

The U.S. Infantry, U.S. Marine Corps Infantry, and U.S. Armor alone account for more than 57% of all U.S. Military combat-related amputees.¹ Approximately 89% of U.S. Military service members with amputations do not return to duty.^1–5 Restoring prosthetic intuitive neural control with somatosensation is a key to improving the safety and efficacy of prosthetic locomotion, which would allow U.S. veterans to continue their military career and be more active in their civilian life. The purpose of the study was to integrate the methods of bidirectional neural interface, previously developed and tested in our animal studies^4–7 into the existing anthropomorphic passive Free-Flow Foot (FFF) and Ankle prosthesis, WillowWood, Mount Sterling, OH. The FFF was chosen because it has the highest Index of Anthropomorphicity among lower-limb prostheses and was the first anthropomorphic prosthesis brought to market.^8,9

Background

The optimal lower-limb neuroprosthesis has the following characteristics¹⁰: It places minimum pressure on the residuum; it incorporates bidirectional control; it allows the amputee to experience somatosensory feedback; it is light; it operates with low-energy consumption.

The NIH, DoD, and Veterans Administration have invested significantly in the control and sensory perception of prosthetic devices and bidirectional neuroprostheses, which has led to promising developments.^11–13 Nonetheless, natural somatosensory feedback from lost lower limbs has not yet been incorporated into any commercial prosthesis.^14–16 Restoring intuitive neural control with somatosensation has the potential to improve the safety and functionality of prosthetic locomotion. For that, it is important to provide the user not only with tactile sensations by electrical stimulation of cutaneous afferents^17,18 but also proprioceptive information from muscle length–sensitive spindle afferents and muscle force–sensitive golgi tendon organ afferents.^11,19,20

The Current Market

The powered ankle–foot prosthesis market is currently immature. A 2021 review of 159 publications on 94 powered ankle–foot prostheses developed in the last 2 decades revealed that only 3 of them were commercially available,¹⁶ with only “Össur Proprio Foot” and “Ottobock Empower Ankle” (previously BiOM)^21,22 considered successful. The 2 prostheses, while the current leaders, have significant limitations. The Össur Proprio Foot’s performance is equivalent to a passive spring foot²³ because its electric actuation adjusts the ankle angle only in the swing phase but is locked during stance. The Ottobock Empower Ankle is more advanced, generating a power impulse during the stance phase as well, but its performance is suboptimal since it does not transmit signals from the wearer’s neuromuscular system.

The advances are even slower for patients with osseointegrated prostheses via direct skeletal attachment (DSA). Despite the efforts of many research groups and work on powered upper-limb DSA prosthetics,^24–28 the idea has not yet been implemented in lower-limb prosthetics.²⁹

One of the main root causes is the risk of falling of end-users and breaking the implant in the bone because of the possible lack or misfunctioning of the control system.

Innovation

This article presents a concept of human-centered bidirectional lower-limb neuroprosthesis, where the feedback is delivered to the nervous system of the user who controls the active prosthetic joints. This approach is distinct from the concept of machine-centered bidirectional neuroprosthesis, where the feedback is delivered to the microcontroller (machine) controlling the active joints.³⁰ Next, it discusses requirements for the foot–ankle prosthesis to be a biotechnological platform for the human-centered bidirectional lower-limb neuroprosthesis able to best utilize its advantages, chief among them the criteria of anthropomorphicity.^8,9

METHODS

Ethics Review Statement

All surgical and experimental procedures were conducted in agreement with the Principles of Laboratory Animal Care (NIH publication no. 86-23, revised 1985) and approved by the Institutional Animal Care and Use Committee of the Georgia Institute Technology (protocols 16089 and 100013) and T3 Laboratories (protocol GT27F) and DoD Animal Care and Use Review Office committee.

Study Design

Our approach was to develop and test the user-centered system for controlling the timing and the volume of the impulse generated by prosthesis motor. The foot and ankle prostheses were selected from among existing devices that met the moment criterion of anthropomorphicity³¹ and which had the maximal corresponding Index of Anthropomorphicity, further described in the section “Foot–Ankle Prosthesis as Biotechnological Platform for Integration With the Developed Control System.”

Bidirectional Intuitive System of Power Control

We developed and tested a sensorized powered transtibial prosthesis for a cat with a novel system of bidirectional intuitive power control and investigated the effects of bidirectional prosthetic control on locomotor mechanics; these results have been partially published.^4,5,7,32 Electrodes implanted into the residual nerves and muscles communicated with an external powered prosthesis, as demonstrated in the animal studies.^{4–6,32–35} The studies were conducted in cats with osseointegration using the skin and bone integrated pylon (SBIP) technology with peripheral neural interface (SBIP-PNI), following the first ever transcutaneous transmission of nerve signals via the central canal of an implant hosted in the residuum bone,³⁶ thus realizing the vision expressed by the originator of DSA, P.-I. Brånemark.³⁷ Bidirectional transmission of the signals between residual nerve and muscle and the powered sensorized prosthesis was achieved via electrode leads passed through a channel in the original porous titanium pylon (SBIP-PNI) implanted into the residuum. Figure 1A shows the initial stage of fabricating the SBIP-PNI, with molds (a) for sintering porous cladding around solid inserts (b) with canals for wires between the residuum and outside prosthesis. ^{Figure 1B} shows the SBIP-PNIs after sintering. White dotted lines show the orientation of the central channel; zone (a) is designed to be implanted into the marrow canal of the residuum bone; zone (b) is surrounded by the residuum skin; zone (c) is an outside part of the implant used for the attachment of the prosthesis.

FIGURE 1.

Skin and bone integrated pylon with a channel inside for peripheral neural interface (SBIP-PNI) fabricated for the cat study with a powered sensing prosthesis having bidirectional control. Figure 1A illustrates fabricating the SBIP-PNI with molds (a) for sintering porous cladding around solid inserts (b) with canals for wires between the residuum and outside prosthesis. Figure 1B shows the SBIP-PNIs after sintering. Dotted lines show the orientation of the central canal; zone (a) is designed to be implanted into the marrow canal of the residuum bone; zone (b) is surrounded by the residuum skin; zone (c) is an outside part of the implant used for the attachment of the prosthesis. Figure 1C demonstrates an X-ray image of the right tibia with the SBIP-PNI (upper L-shaped ine), the nerve cuff electrode on the residual distal tibial nerve, and electrode leads exiting the implant. The tibia with implants was extracted from the animal after completion of the study 29 months after implantation. Bottom panels demonstrate results of histological analysis of implant integration with bone and skin 29 months after implantation. Bottom left panel (Fig. 1C Section A): Longitudinal section through the residual tibia and implant hemotoxylin and eosin staining, magnification 4×. Dark arowheads = continuous layer of epithelium over fibrous stump bridging to the edge of the titanium implant; asterisks = bone ingrowth within the implant immediately beneath the skin and continuing along the implant shaft, light arrow = minimal granulation tissue adjacent to emergence of the pylon, clear arrowheads = serocellular debris and crusts within the titanium adjacent to the emergence of the pylon (green box). Bottom right panel (Fig. 1CSection B): Cross-section through the residual tibia and implant hemotoxylin and eosin staining, magnification 10%. Asterisks = extensive mature bone ingrowth and apposition along and within the porosity of the titanium implant (T); P = pylon.

We implanted the SBIP-PNI into distal tibia, electromyographic (EMG) electrodes into the residual gastrocnemius (GA) muscle,³⁸ and nerve cuff electrodes on the distal tibial and sciatic nerves.^5,39 The electrode leads were passed through a small hole in the tibia and then through the channel inside the implant and secured in a connector inside an aluminum box attached to the external part of the implant (Fig. 1B-C). The signal was sent via a wire run through a canal in the SBIP-PNI pylon. Bidirectional transmission via SBIP-PNI was then achieved. A pressure transducer on the bottom of the prosthetic foot triggered stimulation of the residual distal tibial nerve, activating cutaneous and proprioceptive afferents from the missing bottom of the foot. The nerve stimulation parameters (trains of 200 μs biphasic rectangular pulses at 100 Hz for 500 ms duration, strength 1.2× of sensory threshold)⁷ produce realistic tactile perceptions in human amputees¹⁷ and normalize gait in cats with anesthetized ipsilateral paws.⁷ The prosthetic foot pressure signal alone or in combination with EMG signals from the residual soleus muscle initiated prosthetic ankle extension by the actuator.⁷ The cat walked on a walkway oriented horizontally, upslope (27°) and downslope (−27°) with no cutaneous feedback from the foot in the control condition and with stimulation of residual distal tibial nerve during the stance phase. Full-body kinematics and ground reaction forces were recorded by a 6-camera motion capture system (Vicon, UK) and 3 force plates (Bertec Corporation, USA) embedded in the walkway. The ankle resultant moment was computed using inverse dynamics analysis during locomotion before and after prosthesis implantation.^40,41 After completion of experiments 29 months post implantation, the residual limb with the SBIP-PNI pylon was surgically removed for histological analysis in a survival surgery under aseptic conditions and general isoflurane anesthesia. After recovery from surgery, the animal was adopted by an approved owner. The extracted implant was fixed in formalin and sent to Alizee Pathology, LLC, USA, for histological analysis.

We tested the performance of the developed prosthesis (the resultant ankle moments) in 2 modes of operation: (1) Pressure mode (mode 0), in which the linear actuator extended the prosthetic ankle during contact with the ground and flexed it during swing and (2) EMG mode (mode 2), in which the EMG signal was used in addition to foot pressure signal to control ankle extension during stance while ankle flexion was performed during swing. Each of the above modes of prosthetic operation was tested with and without electrical stimulation of the residual distal tibial nerve. In total, we recorded 534 cycles of level, downslope, and upslope prosthetic walking. To test the statistical significance of the effects of independent factors—Modes of prosthetic operation (intact limb, prosthetic mode 0, prosthetic mode 2), distal tibial nerve stimulation (stimulation on, stimulation off), and slope (downslope, level, upslope), on the peak of ankle joint moment (dependent variable), we used a linear mixed effect model analysis (MIXED, SPSS 19, IBM SPSS, USA). We performed a post hoc paired comparison using the Games–Howell test. The significance level was set at 0.05.

Foot–Ankle Prosthesis as Biotechnological Platform for Integration With the Developed Control System

For better simulation of ballistic synergy seen in normal gait in the new bionic prosthesis, we selected the FFF and Ankle passive system.^42,43 This prosthesis was developed with the aim of meeting the moment criterion of anthropomorphicity, replicating the moment-ankle dependency generated by the ankle joint in the norm.³¹ In patients with typical socket-suspended attachment, it resulted in lower loads on the residuum compared to the most widely prescribed existing prosthetic feet.^44–47 In an observational case–control biomechanical study with transtibial amputees fitted with osseointegrated prostheses, FFF generated the least breaking moment to the abutment (25% reduction) and more propulsion on the anteroposterior axis as well as less internal external and lateral rotation compared to the Triton and TRIAS feet.^9,48 In addition, the rolling joint mechanism of the passive FFF requires only a minimal power supply at the end of the stance period of gait, when it acts as a supplement for facilitating a faster pace or walking on a slope.

The integration between the FFF and novel system of bidirectional control with socket-suspended transtibial prostheses and osseointegration-type transtibial prostheses comes from surface skin electrodes along the interface between the prosthesis and the user. In the motor (efferent) pathway, a linear actuator was activated based on EMG from the residual GA and tibialis anterior muscles. In the sensory (afferent) pathway, electrical stimulation was applied to a branch of the tibial nerve in the residual soleus during early stance, to deliver sensory feedback via surface electrodes. This stimulation is triggered by foot contact with the ground recorded by a pressure sensor. The soleus stimulation also activates the soleus muscle fibers and motor fibers of the residual soleus branch of the tibial nerve, but that activation has no mechanical effect on the prosthetic ankle because the soleus distal portion and tendon are amputated.

A built-in microprocessor in the control electronics process both EMG and pressure sensor output and generate control signal for the linear actuator and electrical stimulation.

RESULTS

Our animal experiments informed the development of the first anthropomorphic lower-limb neuroprosthesis with intuitive bidirectional control and sensation. Selected results of histological analysis of the SBIP-PNI implant with the residual tibia (Fig. 1) demonstrated that the implant can provide infection-free integration with bone and skin for the period of over 2 years. Longitudinal and axial sections through the implant and tissues demonstrated skin and bone ingrowth within the porosity of the titanium implant (Fig. 1). Selected results of the biomechanical analysis of cat locomotion with the sensorized powered transtibial prosthesis are shown in Figure 2. The signal from the pressure transducer on the bottom of the prosthetic foot during the stance phase triggered the onset of ankle extension by the linear actuator in mode 0 and initiated electrical stimulation of the residual distal tibial nerve to inform the animal of the contact with the ground.⁷ In mode 2, the ankle extension started when both foot pressure and EMG signal from the residual soleus muscle were present at the same time (Fig. 2A). Patterns of the ankle extension moment during prosthetic walking in mode 0 and mode 2 were qualitatively similar to the ankle moment pattern before surgery, although the peak moment amplitudes were lower (Fig. 2B). Statistical analysis revealed significant effects (P < .05) of independent factors—Modes of prosthetic operation, nerve stimulation, and slope on the peak ankle extension moment. Specifically, peak ankle extension moment was lower in both prosthetic modes than in the intact limb before surgery in all slope conditions (F_2,504 = 4.7-22.1, P < .009; Fig. 2C). Prosthetic operation mode 0 produced a larger moment peak than mode 2 during upslope walking without nerve stimulation (P = .001; Fig. 2C, left panel). Nerve stimulation affected modes of prosthetic operation differently in different slope conditions; in downslope, nerve stimulation did not have significant effects on the peak prosthetic ankle moment (P = .695), increased the peak moment in mode 0 compared to mode 2 in level walking (P = .004), and had the opposite effects on the peak moments during upslope walking (P = .001; Fig. 2C, right panel). The peak ankle extension moment in intact and prosthetic conditions increased with increasing slope from downslope to level and to upslope walking (F_2,504 = 791.3, P < .001; Fig. 2C).

FIGURE 2.

Selected results of animal prosthetic locomotion experiments. (A) Examples of rectified low-pass (50 Hz) electromyogram of the residual soleus muscle and pressure signal from the bottom of the prosthetic foot during cat level walking (mode 2, no stimulation). (B) Mean ankle joint moment patterns during level intact and prosthetic walking in modes 0 and 2 with and without stimulation of the residual tibial nerve. The mean patterns were obtained by averaging between 6 and 12 walking cycles. (C) Results of statistical analysis of effects of prosthetic mode, slope of the walkway, and stimulation of the residual tibial nerve (see text for details). * indicates statistical difference (P < .05) from the intact condition; # indicates statistical difference between the prosthetic modes.

We have developed the first anthropomorphic lower-limb neuroprosthesis with intuitive bidirectional control and sensation. The prosthesis, called Bionic FFF, combines the outcomes of 2 technologies. One is the passive FFF, the first prosthesis designed to replicate the stiffness of an anatomical ankle joint.⁸ The second technology was developed to evoke tactile sensation of ground contact by electrical stimulation of the sensory nerve^5–7 and to use EMG activity of the residual ankle muscles to control the ankle actuator during plantarflexion at the end of the stance period. A special procedure has been developed to individually adjust the control system to the neuromuscular conditions of the amputee.

We developed the control system for future human trials. Control (efferent) signals from the user to the prosthesis actuator are to be captured from EMG of residual ankle plantar- and dorsiflexor muscles, specifically the GA and tibialis anterior. Sensory feedback (afferent) information about contact of the prosthetic foot with the ground obtained from a pressure sensor located on the foot bottom is reported to the user’s nervous system by transcutaneous electrical stimulation of the residual soleus that contains a branch of the tibial 1 nerve with proprioceptive and cutaneous afferents innervating ankle extensors and skin on 2 plantar surface of the foot. Bipolar surface electrodes placed on the belly of residual ankle flexor and extensor muscles relay recorded EMG to the Bluetooth Low-Energy (BLE) module (mounted on the socket or the prosthesis’ pylon), wirelessly sending the data to the control electronics (mounted on the linear actuator).

The control electronics are designed to generate plantarflexion of the powered ankle joint based on the EMG data, by controlling a linear actuator located roughly where the Achilles’ tendon would be. The actuator actively pulls its arm for plantarflexion and is turned off during the dorsiflexion for the FFF to release energy (Fig. 3). The BLE module delivers sensory feedback via stimulation of the residual tibial nerve, based on the signals from the built-in pressure sensor. For stimulation of the residual tibial nerve, we use bipolar surface electrodes placed over the 13 residual soleus muscles. The electrode placement is selected for each participant based on 14 perception and noxious thresholds (Supplementary Fig. S1). This intensity of stimulation activates groups I, II, and III muscle and cutaneous afferents responsible for tactile and proprioceptive sensations and for evoking spinal locomotor reflexes.^49–51 For stimulation of the residual soleus muscle, we use bipolar surface electrodes placed over the residual soleus muscle. The electrode placement is selected for each participant based on perception and noxious thresholds.

FIGURE 3.

Structure of the Bionic Free-Flow Foot (FFF) control. (A) Passive first plantar flexion after “heel-on,” (B) ankle joint is passively dorsiflexed, cushioning is compressed, and the stored energy is released during the second plantar flexion, (C) ankle joint is in transition to second plantar flexion by both stored energy in the FFF and the linear motor, and (D) ankle joint is actively plantarflexed by the linear motor. Cushioning is compressed, and the stored energy is released during the dorsiflexion of the swing phase.

The built-in multipoint control unit has wireless communication capability, allowing it to communicate with an external processor and update stimulation parameters in real time.

Hardware Specifications

Electromyographic signals are amplified with the INA326 neural amplifier (Texas Instruments, USA) and digitized by an nRF52832 (Nordic Semiconductor, Norway), integrated in the BLE module (Flora Wearable Bluefruit LE Module, Adafruit, USA). Biphasic current stimuli are generated by a constant-current biphasic pulse generator, based on the control signal generated from the nRF52832. The nRF52832 also supports wireless data communication with the external computer system via BLE, to send EMG data to outside and receive stimulation parameters from outside. For the active ankle joint operation, a linear actuator (L16-p-63-12, Actuonix Motion Devices, USA, 168 mm long in total, with the motor arm 0-100 mm long) is integrated with the passive FFF. The actuator’s arm is pulled (shortened) for the active second ankle plantarflexion of the prosthetic foot before take-off. Li-polymer battery and the control electronics are mounted onto the body of the linear actuator (Supplementary Fig. S2). The control electronics generate control signal for the linear actuator, based on the EMG data wirelessly received from the BLE module. The actuator has a back-drive capability and is turned off during the period of passive movement driven by the FFF energy releasing mechanism. The Li-polymer battery has a 4 Ah capacity with 7.4 V nominal voltage output.

The system current consumption is 660 mA at the maximum torque of 10 Nm (650 mA for actuator and 10 mA for electronics), and the system current consumption at passive operation (back drive) of the motor is 10 mA (only for electronics). The usage time is estimated at ∼8.16 hours, assuming 75% duty factor of the actuator action. The selected linear actuator can generate torque up to 10 Nm. This is 25% of the required torque at the late stance (∼0.5 Nm/kg, or ∼40 Nm for an 80 kg adult). The estimated weight of the Bionic FFF is 0.93 kg, composed of the FFF (0.72 kg), linear actuator (0.1 kg), electronics (0.01 kg), and 2-pack 4 Ah battery (0.1 kg). The weight of the Bionic FFF is 42% of Ottobock Empower’s (2.2 kg), a commercial transtibial prosthesis with an active ankle joint operating at ankle plantarflexion.

DISCUSSION

Considering the growing attention paid to developing functional powered prostheses with various specifications of their control systems, it is important to compare them objectively. The summary of such a comparison (Table I) clarifies a main distinction of the systems in the market and in research with the proposed Bionic FFF prosthesis. Following the study in cats, we aim to achieve similar results in human subjects. Because of its desirable anthropomorphic properties, we will apply a bidirectional neural interface to control the passive FFF. However, in a contrast with the cat study, in which wires running through an osseointegrated percutaneous implant directly stimulated a sensory nerve and recorded muscle activity, stimulation and recordings in a first human trial will be accomplished through surface electrodes. This approach does not require an additional surgery in the subjects with osseointegration but allows for detailed analysis of the new algorithm and hardware. Only after successful completion of this trial, we would apply for permission to implant the nerve cuff electrodes to the residuum.

TABLE I.

Comparison of the Current Powered Neuroprostheses and Proposed Bionic FFF

	Powered prostheses	Powered prosthesis with system feedback	Powered prosthesis with bio-feedback—Bionic FFF
Trigger point	Ankle/knee	Ankle/knee	Ankle
Activation mechanism	EMG-based/motion-based	EMG-based/motion-based	EMG-based/motion-based
Sensing mechanism	n/a	Plantar pressure/joint torque	Plantar pressure/joint angle
Feedback mechanism	n/a	Feedback to the system	Feedback to the user
Control mechanism	User control	Collaboration between user and built-in system	User control
Overall characterization	Unidirectional neuroprosthesis	Bidirectional machine-centered neuroprosthesis	Bidirectional human-centered neuroprosthesis

The overall characterization of the prototype is a bidirectional human-centered neuroprosthesis, where the feedback is delivered to the nervous system (human) and the human controls the active joints versus the concept of machine-centered bidirectional neuroprosthesis.³⁰ The design of the prototype is informed by the results of this and our previous cat study.⁷ The cat animal locomotion model has important advantages over other animal models. The cat is a relatively large cursorial animal moving on substantially extended legs as humans, as opposed to noncursorial animals (e.g., rodents) that locomote on strongly bent legs. The mass of cat body segments is large enough so that cats, as humans, use segment inertia to extend and flex some leg joints passively during swing; this is not the case for rodents.²⁰ Since many mechanical variables (ground reaction forces, joint moments, and power) for cats⁵² and humans,⁵³ once expressed in percentage of the body mass, look very similar, we anticipate that the bidirectional neural interface successfully trialed on cats can be extended to humans and will maximize the efficacy and safety of powered locomotion.

CONCLUSION

1. Two technologies have been presented in this article: The technology of bidirectional neural interface and the technology of the anthropomorphic passive FFF and Ankle prosthesis.

2. Integration of both technologies is suggested for conducting conclusive clinical trials.

3. Positive outcomes of future clinical trials would usher the development of a powered bionic prosthesis with intuitive control and sensation (Bionic FFF).

SUPPLEMENTARY MATERIAL

Supplementary material is available at Military Medicine online.

FUNDING

This study was supported in part by the Grant W81XWH-16-1-0475, U.S. Department of Defense Orthotics and Prosthetics Outcomes Research Program; Grant R44AR43290 National Institute of Arthritis and Musculoskeletal and Skin Diseases & National Center for Medical Rehabilitation Research - National Institutes of Health; U.S. Department of Defence RESTORE Award W81XWH2110215- DM190659, 2021 Bionics Queensland Challenge Major Prize – Mobility and the 2024 NHMRC Idea grant 202963.

SUPPLEMENT SPONSORSHIP

This article appears as part of the supplement “Proceedings of the 2023 Military Health System Research Symposium,” sponsored by Assistant Secretary of Defense for Health Affairs.

CONFLICT OF INTEREST STATEMENT

None declared.