3D Printed Motor-Sensory Module Prototype for Facial Rehabilitation

Abstract

This work demonstrates the first 3D printed wearable motor-sensory module prototype designed for facial rehabilitation, focusing on facial paralysis. The novelty of the work lies in the fast fabrication of the first fully soft working prototype, including feedback control, with a focus on the methodology for individual customization. Facial paralysis results from a variety of conditions, and more wearable and modular technologies are needed to address the complexity of facial movement rehabilitation. Smiling muscles are especially important for both expression and eating, and so this work focuses on this motion as an example of how the module can be applied to mimic and support needed muscle movement. A generalized actuator-sensor pair with a feedback control system is created to translate signals from smiling on the healthy side of the face (notably temporal and zygomatic branch) to actuation on the paralyzed side of the face for augmented physiotherapy. Fabric and a sensor fluid are integrated during the silicone printing process to create a multicomponent wearable that is ready to use with minimal postprocessing. The actuators' force and vertical contraction results under a 0.98 and 1.96 N load meet the 1–7 N requirements needed for smiling. It is a challenge to measure soft surface-based force and contraction ratio consistently; therefore, a novel modular surface is designed to simulate the interaction of skin and bone using 3D printed hard plastic (bone) and a silicone sheet (skin). The actuator is tested on top of four different repeatable and standardized surface morphologies, and results reveal that the actuator force application will vary based on topography and hardness of the facial surface. Demonstration of the complete system on the face while collecting sensor and pressure data serves as a proof-of-concept and motivates potential applications in rapid customization of highly specialized soft wearable orthotics, prosthetics, and rehabilitation devices. This unique actuator-sensor combination can have additional applications for wearables due to the (1) customizability, (2) closed-loop control, and (3) unique “grounding” test platform.

Objective

Wearable robotics and their customization are two frontiers for development toward rehabilitative technologies. This field holds immense potential for human augmentation and increased quality of life. Research into exosuits,^1,2 gloves,^3–6 and joint rehabilitation technologies^6–11 is reflective of a profound need for maintaining quality of life by keeping the human body functioning healthily for as long as possible, while reducing mental and physical pain in the process. For soft wearables specifically, improved manufacturing methods are explored to make the production highly customizable for individual patients and their medical conditions. Multicomponent and multi-material 3D printing are particularly well-suited manufacturing methods for wearables because of the possibility to integrate diverse and multiple soft elastomers and conductors for actuation and sensing in the same device.

Facial paralysis (palsy) introduces asymmetric expressions that are embarrassing¹² and physically restricting for the patient. One of the most impactful facial movements is the smile, and so loss of the smile causes distress due to loss of the muscles used in eating, drinking, and maintaining a “normal” facial appearance. Reasons for facial paralysis can stem from Bell's Palsy,¹³ stroke, tumors, or other neurological conditions.¹⁴ Prevalent paralysis symptoms are the drooping of the eyebrow and the corner of the mouth as well as difficulty closing the eye and the mouth due to the affected facial nerves.^13,15 Restoring facial movement is key for ensuring that those with the condition do not suffer embarrassment due to their asymmetrical expression and are able to eat, drink, and speak normally.¹⁶

Along with prescribed medicines, facial muscle exercises have shown promise for improving the condition,^13,17 including cases where patients need further rehabilitation after not fully recovering from Bell's Palsy.¹⁸ Facial movement rehabilitation practice can consist of enhancing desired facial muscle movements or reducing unwanted facial muscle movements.¹⁹ Physiotherapy for this condition involves exercises performed using the palsied side of the face, such as smiling, which mimic the healthy side to achieve symmetry.^16,20

Although there is much research into rehabilitative wearables for exosuits and joints, there is a gap in the understanding of how soft devices can support facial movement. This gap is likely due to the complex nature of facial movement, the size limitations of current actuator solutions, and the lack of diagnostic tools that give consistent and quantifiable measurements. There is a high degree of variability between human faces,²¹ which makes use of a generic size difficult.

The few recent facial actuator solutions currently focus on restoring motor function of the face when one side has been paralyzed.^22–26 Rehabilitative facial device prototypes exist that rely on cable-driven/shape memory alloy hard helmet systems,²² and shape memory alloy helices.^23,24 These devices have also been used as training systems for rehabilitation.^25,26 However, some previous works rely on the pulling of wires from a single source moving across the face.^22–24 These devices are not modular enough to enable complex force application in areas of the face where cables could overlap or press into delicate skin. Other previous works employ hard origami fingers requiring complex manufacturing, preventing easy customization.^25,26

Achieving symmetry for those with palsy is critical to reducing pain and stress levels, but it is difficult to achieve during healing. The challenge is in producing simultaneous expressions on both sides of the face when there is no fixed or target motion that is the same for all patients. The ability to sense the motion of the healthy side of the face and translate that to identical movement in the paralyzed side of the face would augment and potentially increase the effectiveness of facial physiotherapy. Informational feedback from the wearable to the user can also be a valuable tool for sustaining motivation.²⁷ In addition, if the wearable has the capability to connect with game-based rehabilitation exercises, the potential for patient adherence to the rehabilitative protocol is increased.²⁸

A wearable device designed to encourage symmetry via existing healthy facial movements can help improve patient outcomes. For this type of device, several tasks need to be achieved: (1) Facial movement is measured and quantified, (2) the device actuates, (3) the device enables alignment with muscles for optimal actuation in the correct direction, and (4) manufacturing outputs a customizable and cohesive system. Preferably, this wearable has dimensions based on individual facial feature sizes as well as customizable actuation behavior based on the severity of the palsy condition. No diagnostics currently have these qualities, as current physiotherapy relies on coarse and subjective methods. Actuators and sensors currently exist that could fulfill these needs, but no comprehensive solution exists.

In this work, a soft facial wearable motor-sensory module prototype containing a paired vacuum actuator and an embedded strain sensor is created via multicomponent 3D printing with silicone (Fig. 1). This article outlines the first trials of the working prototype and details the comprehensive methodology for the prototype's customization per individual rehabilitative needs.

FIG. 1.

Photograph of the sensor-actuator pair module on a healthy subject. The 3D printed actuator-sensor module enables movement on one (healthy) side of the face to be mirrored on the side of the face affected by paralysis to help patients retraining muscle movements during physiotherapy. The 3D printability increases options for customizability not only in size but also in shape and force application, and soft material components increase the conformability of the device to a variety of face shapes. The pairing of a sensor-actuator system with a feedback loop also enables potential tracking of patient progress. Color images are available online.

A generalizable actuator-sensor pair with accompanying feedback control is characterized for support of a smiling motion (corner of the mouth uplift) with actuator placement supporting zygomatic branch muscle actuation. The module purpose focuses on facial physiotherapy. The novelty in this work lies in the system-level prototype creation. The actuator is characterized for force and deformation under vacuum, and force behavior over four types of irregular curvatures covered in a skin-like membrane. The sensor is characterized for resistance under strain. As a paired system, the sensor reads strain from the healthy side of the face and the circuit translates that strain into contraction force from the actuator.

The presented actuator can contract with 1–2 N of force (in the range of facial muscle^26,29). A demonstration on a healthy subject's face validates that the prototype module can recreate the smiling motion of the face by mimicking the same muscle movement of the opposite side. Lastly, several actuator examples are printed to demonstrate customizability of the module via 3D printing.

It is still difficult to invent a comprehensive “treatment course” that is suitable for every facial palsy patient. Therefore, we based our first proof-of-concept based on the actual physiotherapy session (which also varies greatly between the patient conditions, progresses, physician, therapist, etc.) Again, our effort was not to create a “treatment” but to showcase the potential capacity of the device. Therefore, we tested the proof-of-concept device on the authors' face (only to illustrate the dimension of the device and its stiffness, and spatial resolution). Also, we have made a specific effort to create a non-uniform, undulated, and customizable testing platform where we show four different “loading” conditions to validate the mechanical performance of the actuated device.

The contributions of novelty in this work are: 1.

The design and prototyping of a complete soft system for wearable actuation and sensing meant for facial physiotherapy.

Novel test and diagnostic experiment design to recreate distinct facial topographies for specific rehabilitation needs.

Addressing soft material manufacturing challenges (3D printing) for wearable robotics.

This work is also transferrable to existing literature and platforms, such as adding sensors to existing actuator designs via 3D printing and using sensors to diagnose and quantify required actuator capacity and dimensions. The next step of this study will involve collaboration with clinicians to (1) further investigate facial movement data collection (healthy and impaired), (2) create a streamlined metric for “healthy” facial movement, (3) based on (1) and (2) design an experimental protocol and platform to conduct clinical studies, and (4) based on (3) design treatment exercises and validate results.

Materials and Methods

The design and fabrication of this work focus on creating the wearable as a complete system. Because of the high level of variability in human faces, the system must be quickly updatable based on improved design iterations. The manufacturing is based in 3D printing, which allows for customization and iteration for anyone having this printer. However, other rapid manufacturing methods can potentially be used. The novelty of this work lies in the novel prototype system itself, so it allows different sensor-actuator combinations to be adapted.

Design for biomechanics

The actuator operational safety is critical, because tissues around the eyes and mouth are sensitive. Vacuum actuation presents a viable option for facial robotics, because the actuators can also be soft, lightweight, and powerful.^30–33 Patterned buckling geometry has previously enabled muscle-like contraction with high forces^31,32 and motion control.^34–36 This safety factor is also needed for strain sensors for the face. The enclosed resistive element needs to be nonhazardous and nonleaking. Low hazard fluids such as glycerol are well suited to this purpose, with resistive yet low hazard particle additives.^37,38 Vacuum actuators also leak on puncture, instead of bursting under positive pressure, thus further increasing safety.

Evaluating a facial wearable module is especially challenging due to the variability in facial features and skin behavior; therefore, the development of a unique test setup is needed. Facial skin, in particular, has many different mechanical properties^39,40 and designing a testing system to approximate these properties, as well as the inherently uneven facial surface, is critical to give a more accurate representation of the performance of the wearable.

Design and fabrication

An advantage of using 3D printing to create wearables is the ability to work within a design loop of creation, manufacturing, testing, and fast iteration. This enables a high level of customizability for both soft and hard components. In this work, one example is shown to illustrate the functionality available for smiling physiotherapy, though many other examples are possible. This process can also eventually be fully automated, requiring no manual assembly. To demonstrate the potential result of automation without needing to augment the current 3D printer system, fabric and hard components are placed by hand. The conductive fluid is also extruded by hand into the sensor reservoir. The 3D printing system was identical to previous work, where thermoset silicone was printed using zero support.⁴¹

The 3D printed actuators were designed to be fully compliant when adhered to the face. These were made out of silicone (Dragon Skin 10 Very Fast with Thi-Vex and Silicone Thinner additives, all from Smooth-On) and polyester mesh fabric on a custom-built 3D printer. The silicone formulation and silicone printing method is detailed in previous work.⁴¹ Materials characterization also informed the manufacturing process (Supplementary Fig. S1). The actuator design (Fig. 2a, b) was based on previous successful vacuum soft pneumatic actuators (SPAs) using rectangular geometry^30,42 to achieve force around 1–2 N needed for facial movement.²⁹

FIG. 2.

Actuator-sensor module prototype. (a) Wire frame of actuator model and (b) resulting printed actuator after adhesion of reinforcing polyester mesh and Velcro in the tab areas. Velcro is only attached to the tab areas for characterization, not for attachment to the face. Those regions that attach to the face are left as-printed. (c) Wire frame of sensor model and (d) resulting printed sensor after attachment of polyester mesh tabs for characterization. Color images are available online.

Because the printing method used continuous extrusion of silicone, thin single-line paths of the 2D actuator shape were drawn in Illustrator, extruded into a 3D model in SolidWorks, and finally sliced in Simplify3D to force the G-code to follow the specific path and ensure that no pathways overlapped. More complex pathways could potentially be developed with custom algorithms.⁴³ The tabs were designed using extruded rectangles in SolidWorks. The length of the actuator (70 mm without tabs) was based on the length from the corner of the mouth to the cheekbone area of the author. The actuator height was 12 mm. This zygomatic region is the principle muscle group responsible for the movement of the corner of the mouth. One-millimeter-thick and 17 mm long silicone tabs were added to both ends for testing and for adhesion to the face, with the tabs centered at both ends to reduce anisotropic contraction.

The actuator and sensor were designed to adhere to a movable soft surface and retain enough strength for repeated use. Three areas of the actuator were reinforced with polyester mesh fabric (Fig. 2a, b). To embed the fabric into the actuator, the first layer of the model was printed and the extruder was then moved away. A piece of polyester mesh was then rolled into the uncured layer, creating a composite, and the print was resumed until the layer just before the tabs was reached. After that section of the print was complete, a second 10 mm wide strip of polyester mesh was lightly pressed into the top of the print to adhere to the two ends and the middle chamber areas. Two hard plastic rectangles were also then put next to the ends of the actuator to serve as support structures for the tabs, and then the print was resumed. The next section was printed all the way up until just before the top encapsulating layers, and then the extruder moved away and a third piece of polyester mesh was pressed lightly into the top of the just printed section, making sure not to have the fabric extend over the print boundary; extending the fabric over the print boundary would have created vias for air to leak. After the third piece of fabric was placed, the top encapsulating layers of the model were printed (see Supplementary Fig. S2 for more details).

About 10 min after the print was finished, the print was fully cured and could be removed from the build plate. Polyester mesh pieces were glued to the tabs and the end surfaces with Sil-Poxy, leaving about a 10 mm portion empty on the underside of each tab for adhesion to the face. The air inlet tube was added by cutting a small slit in one side of the actuator and pressing the silicone tubing in, then gluing it in place with Sil-Poxy. Tubing length ranged from 14″ long for Instron and custom surface testing to 40″ long for the demo, with a 1/16″ inner diameter. Photos of this process are available in the Supporting Materials.

This sensor type was chosen to show how the 3D printing system can enclose both fluids and hard components. The difficulty when creating soft sensors is the incorporation of fluid components as resistive elements. The ideal sensor would not drift and should have repeatable resistance data versus deformation. However, fabricating custom sensors moves the result away from this ideal, especially when fabricating the sensor via a mixture of 3D printing and manual construction. The manual syringing of fluid into the silicone reservoir was not controlled for volume because of the difficulty of overcoming the surface interactions between the carbon/glycerol fluid and the silicone when filling the reservoir, as well as the small size of the fluid reservoir. There are diverse options for embeddable sensors but each have challenges to overcome for wearable devices. Therefore, the solution presented here is meant as an example and should not be considered as the exclusive option.

The sensor models are a hollow dumbbell shape with a cavity for conductive fluid and two electrodes on each end for four-point resistance measurement (Fig. 2c, d). The electrodes were fabricated by rolling the end of each wire into a small spiral (two turns) and then melting solder onto the spiral, forming a small ball. Two electrodes were each melted into an oval-shaped thermoplastic anchor (two in total for each end of the sensor) to help them stay in place after encapsulation. Sensor reservoirs were fabricated by printing most of the dumbbell (all except the top encapsulating layer) in silicone and then moving the extruder away. The anchors were then pressed lightly onto the silicone on each end of the sensor. The wires were pressed lightly through the still-wet silicone ends until about halfway down the model. After about 5 min, the conductive carbon solution was pipetted into the reservoir until the liquid level reached the top of the reservoir (solution composition by weight: 14.3% Carbon Powder [99%, ABCR.de], 85.7% Glycerol [99%, Sigma Aldrich]). Then, the top cap of the sensor was printed to encapsulate the electrodes and the solution.

Module prototype characterization

There are several challenges when developing a testing protocol for wearable technology. Although traditional methods such as blocked force and contraction are good indicators of force application, the actuator's lack of interaction with a surface like the face is a drawback for demonstrating real-life functionality. A surface that can more accurately represent a diverse range of features where the wearable will be attached is important for retrieving more accurate data. Both surface topography and material properties should be emulated in an ideal testing setup. By addressing these needs in a standardized testing platform, a data set more representative of the real-life behavior can be compared with traditional methods. Both traditional and custom tests are used in this characterization.

Actuator contraction was tested at facial forces in the range of 1–2 N and a slow frequency (6 cycles/min) to mimic the time periods possible during physiotherapy sessions. Actuator contraction distance was determined via application of a cyclic vacuum pressure while one end of the vacuum actuator was fixed and the other was connected to either a load of 0.98 or 1.96 N for 10 cycles. The vacuum pressure was applied in 5 s on/5 s off intervals to operate within a time range that was reasonable for complete re-pressurization of the actuator. Actuator contraction data (lift distance versus time) was recorded via video and processed with Tracker software (Tracker specifications are available in Supporting Materials). Actuator blocked force was measured using an Instron tensile tester. The actuator was clamped into the top and bottom grips, and then vacuum pressure was applied in 5 s on/5 s off intervals while force data were collected.

Sensor extension was tested based on approximate facial deformation conditions. Sensor signal response was determined using a custom DAQ connected to LabView 2016 software in a four-point resistance measurement. The sensor extension, 17 mm, was the distance difference between the corner of the author's mouth at rest versus the corner of the author's mouth during a smile (Supplementary Fig. S3). The sensor was tested for 100 cycles with the following protocol for each cycle: Extend to 17 mm at 1000 mm/min, hold for 5 s, release at 1000 mm/min to 0 mm, and hold for 5 s.

A custom testing rig with four separate topographies was developed and prototyped specifically to test the actuators on a surface similar to facial structure and material properties (Fig. 3). This was called the skin and bone modular testing setup (SBMTS). The base was created with a hard plastic 3D printed peg board with modular pegs (Supplementary Fig. S4). The membrane attached to the base was made with a film spreader and consists of a 1 mm-thick Dragon Skin 30 membrane (0.8 wt% Silc Pig White dye, 10 wt% Silicone Thinner, balance Dragon Skin 30 Part A and B in equal amounts) airbrush-painted with a thin layer of Psycho Paint (Smooth-On) (2.8% Silc Pig Red dye, 69.4% NOVOCS Matte solvent, balance Psycho Paint Part A and B in equal amounts) through a laser cut stencil, forming a dot pattern that stretched with the membrane. The outer edges of the membrane were glued with Sil-Poxy to an ∼12 mm wide polyester mesh fabric strip to reinforce the membrane edges. Holes were then cut through the silicone/fabric edge to attach the membrane to the peg board with machine screws and washers. The silicone membranes were made to approximate the deformation of facial skin with a tensile modulus that falls within skin's Young's moduli range. Young's moduli of human skin range from 0.05 to 20 MPa depending on test method.⁴⁰ There are existing tutorials for creating suture practice skins that mimic skin with multiple layers,⁴⁴ but because human skin mechanical performance changes due to age and hydration,⁴⁵ and its overall behavior is complex (anisotropic, nonlinear, and viscoelastic⁴⁶) a simplified single-layer membrane solution was chosen to not overly complicate the experimental setup with unknown correlations.

FIG. 3.

Experimental platform for soft actuator performance. (a) The tested structures of the SBMTS are flat, convex middle, concave vertical, and concave horizontal. The pegs are 5.8, 12.2, and 18.5 mm above the base, making the max difference in height 12.7 mm. Four surface curvatures are created using these pegs: (1) flat (all 5.8 mm pegs), (2) convex middle (center of nine 18.5 mm pegs surrounded by a square of 12.2 mm pegs with the rest as 5.8 mm pegs), (3) concave vertical (column height from left to right as 12.2, 18.5, 12.2, 5.8, 5.8, 5.8, 12.2, 18.5, 12.2 mm), and (4) concave horizontal (same configuration as the concave vertical rotated 90°). (b) The membrane was attached to the surface with machine screws and washers. The actuator was attached at Velcro points on the membrane and hooked up to a pulley leading to a force gauge. SBMTS, skin and bone modular testing setup. Color images are available online.

The SBMTS board was placed into a pulley rig and bolted to the bottom of an Instron. One side of the actuator was hooked to the force gauge through the pulley using a Trilene 100% fluorocarbon fishing line and a small metal hook. Baby powder was applied to the underside of the actuator (touching the membrane) and to the membrane area below the actuator to reduce the effects of friction. To try and decouple the force values from the role of any adhesive attaching the tabs to the membrane, one part of a two-part Velcro was glued with epoxy to the membrane and the other side of the Velcro was glued with Sil-Poxy to the actuator tabs. Both tabs were then attached to the membrane via the Velcro in line with the pulley. The pulley line was then preloaded with 0.5 N of force to hold it taut, and vacuum was applied to the actuator from −86 kPa (for 5 s) to atmospheric pressure (for 5 s) for a few cycles. A video was taken of the actuator next to a ruler, and the amount of extension in the membrane for each of the four surfaces was determined. The extension of the membrane during full actuation on the flat surface was ∼3 mm in 1 s. The concave vertical, concave horizontal, and convex middle extension values measured using the same method were 4, 1, and 1 mm, respectively. The speed of extension on the first sheet (180 mm/min) was used to input into the cyclic membrane calibration tests described next using the separately determined membrane extension distances. To separate the force results of the actuator from those of the sheet, force calibration data were collected before each actuator surface curvature test with the fishing line threaded through two hard plastic tabs with holes, with one tab allowing sliding motion of the line (Supplementary Fig. S5). The tabs were pulled to the extension distances experienced by each membrane surface (determined earlier) and then released for 10 cycles. Each force difference data set (3 total/sheet) was averaged to determine one average force difference per surface type. After calibration, the resulting force for each actuator/silicone surface pair was then recorded versus time in coordination with the opening and closing of the vacuum valve. The calibration values were then added to the actuator data to isolate the force behavior to each actuator alone.

The demonstration of the wearable motor-sensory module prototype was performed on a healthy subject (the author). The actuator was adhered to the skin near the corner of the mouth and the area just above the cheekbone to align with the zygomatic muscle. The sensor was adhered to the center of the lower chin and the corner of the mouth on the other side of the face to capture “healthy” smiling motion. Adhesive (Kryolan Silicone Adhesive Regular Bond) was first applied to the areas of the face where the sensor and actuator were going to be adhered. The adhesive was then applied to the specified areas on the actuator and sensor. Then, the adhesive was left to dry in both areas for 5 min. Once dried, the parts of the wearable covered in adhesive were pressed onto the areas of the skin to set the bond. The actuator was then hooked up to the vacuum circuit, and the sensor was hooked up to the control system. The sensor value when the subject was smiling was quickly determined using the analog signal serial readout, and that value was uploaded with the control code to the Arduino as a threshold value for actuation initiation. A video was recorded of the smiling motion of the “healthy” side translating to the contraction of the actuator/smiling mimic on the “paralyzed” side.

Results

Actuator and sensor characterization

Performance of the actuator under a blocked force and free extension condition is outlined in Supplementary Figure S6. Blocked force data for three actuators over 100 cycles were in a range of about 5–6 N of blocked force assuming no contraction (Supplementary Fig. S6a). Average contraction with a load was about 10 mm for 0.98 N and 8 mm for 1.96 N (Supplementary Fig. S6b). These results are a good start to show that with a 1–2 N load the actuator can pull the corner of the mouth about 10 mm. Sensor data showed drift in the initial data but approached a steady state after a few minutes of use (Supplementary Fig. S6c). The results show that the actuator can successfully operate in the force ranges required for creating a symmetry in muscle contractions to smile, and that the sensor gives a clear enough signal to differentiate a strained versus unstrained state.

SBMTS actuator characterization

SBMTS actuator characterization showed that different underlying hard structures will vary force application in the actuator (Fig. 4). The flat and concave vertical surfaces allowed the actuator to provide more force because of a lack of obstructing geometry pressing into the actuator when compared with the other surfaces. Both the concave horizontal and the convex middle have geometries that press more into the actuator and so obstruct the actuator, even with a low-friction surface. This is useful because although the force values determined are adequate enough to move the mouth (1–2 N), if more force is desired, actuators may have to be designed to apply higher forces for different underlying surface features. Interesting behavior occurred when there was less of an obstructing hard surface underneath the actuator. The actuator in the concave vertical setup pressed into the membrane, because there was an air gap between the membrane and the modular pins. Effectively, it sunk into the loose membrane areas. Overall, this result confirms the assumption that the actuator design needs to consider the topography of the face as well as the underlying hardness or softness of the skin and bone composite surface. Harder surfaces and protruding features (such as the cheekbone) will need an actuator with higher force application than areas with softer surfaces (such as the hollow of the cheek).

FIG. 4.

Actuator testing results on the SBMTS. Four surfaces of different contours: flat, convex middle, concave vertical, and concave horizontal mimic the possible facial structures on which wearables will function. The force results of four actuators tested across the four different surfaces reveal that the force applied by the actuators does change based on surface topography. The flat and concave vertical results showed the most force application due to the lack of geometry pressing up into the actuator. Both the concave horizontal and convex middle have geometrical features that press against the actuator, which, even with a low friction surface, restrict motion.

Motor-sensory module prototype demonstration

To validate the module's performance as a wearable system, we measured the readings on a healthy subject. The operation steps and five cycles of sensing and actuation are shown in Figure 5. The action sequences were performed at irregular intervals to ensure the actuator was not accidentally responding. A section of the video source of Figure 5 is shown in the Supporting Materials (Supplementary Video S1). Green regions show the sensor readings and actuator pressure of the initial relaxed state. Purple regions show the same information for the half smile from the “healthy” side of the face. Orange regions show the same information for the time during actuator contraction. The sensor signal when smiling ranged from ∼575 to 690 kΩ, with a threshold value of 616 kΩ chosen for the beginning of actuation. The time from sensor detection of the threshold resistance to the full vacuum actuation was ∼1.1 s. This time could change based on a different geometry (inner volume) of actuator and the type of tubing used. The data for sensor resistance with strain (dark gray) and pressure (blue) versus time of test show an immediate actuation (decrease in pressure due to vacuum) after the sensor reaches the threshold value (person smiles). There is some signal interference in the circuit (noise after actuation) that causes the baseline of the sensor data to shift upward when the pressure valve signal is sent (Supplementary Fig. S7), but the sensor data still clearly show as a peak before each actuation cycle commences. The system was successfully demonstrated to show its potential applicability for facial exercises.

FIG. 5.

Sensor and actuator reading and performance on the face. The sensor was adhered to the center of the chin and the corner of the mouth. The actuator was adhered to the corner of the mouth and just above the cheekbone, laying in the direction of the zygomaticus muscle region. The sequence of actions for the module were (1) relaxed face (green), (2) smile on the healthy side of the face (purple), and (3) actuated smile from actuator contraction (achieving symmetry) (orange). Actuator gauge pressure data are shown in light blue, and sensor data are shown in gray. The sensor resistance ranged from ∼575 to 690 kΩ during smiling. Interference in the Arduino circuit increased the overall values of resistance when actuation signals were sent, but the peaks in the resistance data still show the increasing strain in the sensor during smiling. The action sequences were performed at irregular intervals to ensure the actuator was not accidentally responding. Color images are available online.

To illustrate the potential of these methods to create more custom geometries, other actuators with varying force characteristics and geometries were 3D printed. Starting with the basic actuator (Fig. 6a), a graded force profile (Fig. 6b), graded contraction profile (Fig. 6c), or a smaller actuator (Fig. 6d) can be created. The 3D printing makes this possible and provides options if different types of contraction are needed. Eventually, 3D printing can produce full masks with multiple force profiles (concept in Fig. 6e). By introducing materials with a higher tensile modulus into this 3D printing system, smaller and stronger actuators can be created. The printed examples outline the larger scope of the module enabled by fast fabrication.

FIG. 6.

Alternate actuator designs. (a) The basic actuator design can be altered to have features customized for specific facial rehabilitation needs. A few simple examples presented here include (b) a graded force profile, (c) a graded contraction profile, and (d) a smaller footprint. The actuator can be adhered in different areas to provide different amounts of force or contraction if needed. Other options include a smaller overall actuator with the same number of chambers, and actuators printed with a tougher elastomer. (e) 3D printing also allows multiple actuators to be printed within the same mask structure if needed. Color images are available online.

Conclusion

A complete and cohesive system is crucial for augmenting physiotherapy exercises, enabling customization in facial devices, and tracking patient progress. The presented novel prototype system initiates this idea via incorporation of both actuation and sensing into a single 3D printed module. We have shown the possibility of this module to be a single printable file customized per patient, such as a mask with sensors embedded into an elastomer base. Because manufacturing in soft robotics is traditionally performed piecemeal with molding and lamination methods, cohesiveness in manufacturing is seldom discussed. The 3D printing advances the production of these wearable robotics via precise control of geometry with possibilities for quick iteration and customization, as well as integration of multiple materials and components. Because this manufacturing method can be fully automated, more complex wearables can be printed with less manual intervention. This unique combination of actuator and sensor can have additional applications for wearables due to (1) customizability, (2) closed-loop control, and the (3) unique “grounding” test platform.

Wearable technologies require a high degree of complexity and a large range of material properties beyond what was described in this article. Manufacturing methods, especially 3D printing, are becoming advanced enough to accommodate multiple materials and multiple components through embedding methods, as seen in this work. The 3D printing creates custom components that can be resized according to the wearer's body dimensions and can also be manufactured relatively quickly. For example, the basis actuator in this work can be made from scratch into a working wearable in about 3 h, including printing, insertion of fabric, insertion of tubing, and adding reinforcing layers. If already created, the wearable installation takes about 15 min.

The production time can be further reduced with improvements to the printing system, such as the automation of component placement with a pick and place machine, and mixing of reinforcing fibers into the silicone while printing. In the context of physiotherapy, this means that the wearable can be created for the patient, tested, altered, and remade all within the same day. The use of an on-body actuator and sensor also eliminates the need for complex facial recognition software or a specialized image capture environment.^47,48 This motor-sensory module can be taken home and used with minimal setup. The actuator and sensor solution developed here is also only one option for those seeking to make a similar system, as the module concept can be applied to other types of actuators, sensors, and fast fabrication methods.

The size of the actuator presented in this work was relatively small (70 × 21 × 12 mm) to reduce the bulk of the actuator on the face, which limited its maximum force output using this 3D printed elastomer. Previous molded shear-based pneumatic vacuum actuators of a similar size (Ecoflex with 00-30 Shore Hardness) have achieved forces as high as ∼5.5 N.³¹ Larger molded vacuum actuators have achieved up to 10 N depending on elastomer material (Elastosil M4601 with 28 Shore A Hardness).³⁰ Stiff pneumatic vacuum actuators made with lamination techniques (plastic films with stiff plastic interior) increased the force of a relatively small (∼100 × 40 × 20 mm) actuator up to ∼29 N (3 kg weight).³² Stiffer yet flexible materials enable higher forces, so 3D printing these types of materials is a priority in future work.

Future work on this project also revolves around the development of wearable solutions addressing other desired muscular motion of the face as well as studying actuator applications for reducing undesired muscle behavior. Further exploration will be performed on how to construct a more complex SBMTS to represent generalized facial movement. Testing of the actuator on additional human faces will be necessary to analyze the behavior on actual patients of several ages and skin types. Using stiffer 3D printable materials for the actuator will increase its future force output. More analysis into the true deformation of human facial skin in relation to adhered facial appliances is needed.

Incorporation of a second sensor into the actuator print itself is planned for coordination of direct strain readings with contraction on human skin. The ability to potentially track patient progress is a valuable tool for quantifying facial symmetry over time. Future collaboration with clinicians will develop this module further, with improvements such as additional data collection techniques, streamlined metrics for facial movement, and experimental protocol for potential treatments. With more development, this novel module prototype can potentially be implemented as a customizable system for facial physiotherapy.

Footnotes

Acknowledgments

The authors would like to acknowledge the help of members of the Reconfigurable Robotics Lab for equipment and general assistance with this project.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This research was supported by the Swiss National Science Foundation through the National Center of Competence in Research (NCCR) Robotics.

Supplementary Material

References

Wehner

, Quinlivan

, Aubin

, et al. A lightweight soft exosuit for gait assistance. In: 2013 IEEE International Conference on Robotics and Automation (ICRA). Karlsruhe, Germany: IEEE,, 2013:3362–3369.

Mengüç

, Park

Y-L

, Pei

, et al. Wearable soft sensing suit for human gait measurement. Int J Robot Res, 2014; 33:1748–1764.

Polygerinos

, Wang

, Galloway

, et al. Soft robotic glove for combined assistance and at-home rehabilitation. Robot Auton Syst, 2015; 73:135–143.

, Kang

, Sin

, et al. Exo-Glove: a wearable robot for the hand with a soft tendon routing system. IEEE Robot Autom Mag, 2015; 22:97–105.

Polygerinos

, Galloway

, Savage

, et al. Soft robotic glove for hand rehabilitation and task specific training. In: 2015 IEEE International Conference on Robotics and Automation (ICRA). Seattle, WA: IEEE,, 2015:2913–2919.

Noritsugu

, Takaiwa

, Sasaki

. Power assist wear driven with pneumatic rubber artificial muscles. In: 2008 15th International Conference on Mechatronics and Machine Vision in Practice, Auckland, New Zealand,, 2008:539–544.

Park

Y-L

, Chen

, Majidi

, et al. Active modular elastomer sleeve for soft wearable assistance robots. In: 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). Vilamoura-Algarve, Portugal: IEEE,, 2012:1595–1602.

Bartlett

, Lyau

, Raiford

, et al. A soft robotic orthosis for wrist rehabilitation. J Med Devices, 2015; 9:030918.

Subramanyam

, Rogers

, Kulesza

, et al. Soft wearable orthotic device for assisting kicking motion in developmentally delayed infants. J Med Devices, 2015; 9:030913.

10.

Park

Y-L

, Chen

, Pérez-Arancibia

, et al. Design and control of a bio-inspired soft wearable robotic device for ankle–foot rehabilitation. Bioinspir Biomim, 2014; 9:016007.

11.

O'Neill

, Phipps

, Cappello

, et al. A soft wearable robot for the shoulder: design, characterization, and preliminary testing. In: 2017 International Conference on Rehabilitation Robotics (ICORR). London, UK: IEEE,, 2017:1672–1678.

12.

, Bundy

, Sadiq

. Psychological distress in people with disfigurement from facial palsy. Eye, 2011; 25:1322–1326.

13.

Holland

, Weiner

. Recent developments in Bell's palsy. BMJ, 2004; 329:553–557.

14.

Facial Paralysis | Conditions & Treatments | UCSF Medical Center. https://www.ucsfhealth.org/conditions/facial_paralysis Accessed June 6, 2018.

15.

Peitersen

. Bell's palsy: the spontaneous course of 2,500 peripheral facial nerve palsies of different etiologies. Acta Otolaryngol (Stockh), 2002; 122:4–30.

16.

Beurskens

, Heymans

. Positive effects of mime therapy on sequelae of facial paralysis: stiffness, lip mobility, and social and physical aspects of facial disability. Otol Neurotol, 2003; 24:677–681.

17.

Baugh

, Basura

, Ishii

, et al. Clinical Practice Guideline: Bell's Palsy. Otolaryngol-Head Neck Surg, 2013; 149:S1–S27.

18.

Lindsay

, Robinson

, Hadlock

. Comprehensive facial rehabilitation improves function in people with facial paralysis: a 5-year experience at the Massachusetts Eye and Ear Infirmary. Phys Ther, 2010; 90:391–397.

19.

VanSwearingen

. Facial rehabilitation: a neuromuscular reeducation, patient-centered approach. Facial Plast Surg, 2008; 24:250–259.

20.

Beurskens

, Heymans

. Mime therapy improves facial symmetry in people with long-term facial nerve paresis: a randomised controlled trial. Aust J Physiother, 2006; 52:177–183.

21.

Sheehan

, Nachman

. Morphological and population genomic evidence that human faces have evolved to signal individual identity. Nat Commun, 2014; 5:4800.

22.

Jayatilake

, Isezaki

, Teramoto

, et al. Robot assisted physiotherapy to support rehabilitation of facial paralysis. IEEE Trans Neural Syst Rehabil Eng, 2014; 22:644–653.

23.

Jayatilake

, Suzuki

. A soft actuator based expressive mask for facial paralyzed patients. In: 2008 IEEE/RSJ International Conference on Intelligent Robots and Systems,, 2008:4048–4053.

24.

Jayatilake

, Suzuki

. Robot assisted facial expressions with segmented shape memory alloy actuators. Int J Mechatron Autom, 2011; 1:224–235.

25.

Firouzeh

. Development of Essential Components for Soft Wearable Technologies. Lausanne, Switzerland: EPFL, 2017.

26.

Firouzeh

, Paik

. Soft actuation and sensing towards robot-assisted facial rehabilitation. In: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017.

27.

Patel

, Asch

, Volpp

. Wearable devices as facilitators, not drivers, of health behavior change. Jama, 2015; 313:459–460.

28.

Hung

Y-X

, Huang

P-C

, Chen

K-T

, et al. What do stroke patients look for in game-based rehabilitation. Medicine (Baltimore), 2016; 95:e3032.

29.

Neely

, Pomerantz

. Measurement of facial muscle strength in normal subjects. Laryngoscope, 2002; 112:1562–1568.

30.

Yang

, Verma

, So

J-H

, et al. Buckling pneumatic linear actuators inspired by muscle. Adv Mater Technol, 2016; 1:1600055.

31.

Yang Dian, Verma Mohit

, Lossner Elton, et al. Negative-pressure soft linear actuator with a mechanical advantage. Adv Mater Technol, 2017; 2:1600164.

32.

, Vogt

, Rus

, et al. Fluid-driven origami-inspired artificial muscles. Proc Natl Acad Sci, 2017; 114:13132–13137.

33.

Robertson

, Paik

. New soft robots really suck: vacuum-powered systems empower diverse capabilities. Sci Robot, 2017; 2:eaan6357.

34.

Yang

, Mosadegh

, Ainla

, et al. Buckling of elastomeric beams enables actuation of soft machines. Adv Mater, 2015; 27:6323–6327.

35.

Ainla

, Verma

, Yang

, et al. Soft, rotating pneumatic actuator. Soft Robot, 2017; 4:297–304.

36.

Verma

, Ainla

, Yang

, et al. A soft tube-climbing robot. Soft Robot, 2017; 5:133–137.

37.

Arabagi

, Felfoul

, Gosline

, et al. Biocompatible pressure sensing skins for minimally invasive surgical instruments. IEEE Sens J, 2016; 16:1294–1303.

38.

Frutiger

, Muth

, Vogt

, et al. Capacitive soft strain sensors via multicore-shell fiber printing. Adv Mater, 2015; 27:2440–2446.

39.

Takema

, Yorimoto

, Kawai

, et al. Age-related changes in the elastic properties and thickness of human facial skin. Br J Dermatol, 1994; 131:641–648.

40.

Pawlaczyk

, Lelonkiewicz

, Wieczorowski

. Age-dependent biomechanical properties of the skin. Adv Dermatol Allergol Dermatol Alergol, 2013; 30:302–306.

41.

Walker

, Daalkhaijav

, Thrush

, et al. Zero-support 3d printing of thermoset silicone via simultaneous control of both reaction kinetics and transient rheology. 3D Print Addit Manuf, 2019; 6:139–147.

42.

Robertson

, Paik

. Low-inertia vacuum-powered soft pneumatic actuator coil characterization and design methodology. In: 2018 IEEE International Conference on Soft Robotics (RoboSoft), Livorno, Italy,, 2018:431–436.

43.

Zhao

, Chen

, Gu

, et al. Connected fermat spirals for layered fabrication. ACM Trans Graph, 2016; 35:1–10.

44.

Creating a Silicone Suture Pad. Smooth-On, Inc. https://www.smooth-on.com/tutorials/creating-silicone-suture-pad. Accessed June 7, 2018.

45.

Kuilenburg J van, Masen

,Heide E van der. Contact modelling of human skin: what value to use for the modulus of elasticity?. Proc Inst Mech Eng Part J J Eng Tribol, 2012; 227:349–361.

46.

Larrabee

. A finite element model of skin deformation. I. Biomechanics of skin and soft tissue: a review. Laryngoscope, 1986; 96:399–405.

47.

Duc

, Minh

. Your face is not your password face authentication bypassing lenovo–asus–toshiba. Black Hat Brief, 2009; 4:158.

48.

Mahmood

. Face Detection by Image Discriminating. Thesis, . Sweden: Blekinge Institute of Technology, 2006.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

4.43 MB

0.49 MB

0.74 MB

0.06 MB

0.18 MB

0.26 MB

1.14 MB

1.62 MB

33.02 MB

0.00 MB