Compliant Buckled Foam Actuators and Application in Patient-Specific Direct Cardiac Compression

Abstract

We introduce the use of buckled foam for soft pneumatic actuators. A moderate amount of residual compressive strain within elastomer foam increases the applied force ∼1.4 × or stroke ∼2 × compared with actuators without residual strain. The origin of these improved characteristics is explained analytically. These actuators are applied in a direct cardiac compression (DCC) device design, a type of implanted mechanical circulatory support that avoids direct blood contact, mitigating risks of clot formation and stroke. This article describes a first step toward a pneumatically powered, patient-specific DCC design by employing elastomer foam as the mechanism for cardiac compression. To form the device, a mold of a patient's heart was obtained by 3D printing a digitized X-ray computed tomography or magnetic resonance imaging scan into a solid model. From this model, a soft, robotic foam DCC device was molded. The DCC device is compliant and uses compressed air to inflate foam chambers that in turn apply compression to the exterior of a heart. The device is demonstrated on a porcine heart and is capable of assisting heart pumping at physiologically relevant durations (∼200 ms for systole and ∼400 ms for diastole) and stroke volumes (∼70 mL). Although further development is necessary to produce a fully implantable device, the material and processing insights presented here are essential to the implementation of a foam-based, patient-specific DCC design.

Introduction

Compliant actuators, relative to their rigid counterparts, uniquely have an increased potential for safe human–robot interaction due to their infinite passive degrees of freedom actuated at low mechanical stresses, a high number of possible degrees of freedom of motion, and an ability to adapt their shape to changing environments.¹ This class of actuators includes piezoelectric polymers,² electroactive polymers,³ dielectric elastomer actuators,⁴ thermo- and chemically responsive gels,⁵ and fluidically pressurized elastomer actuators.⁶ Of these, fluidically pressurized elastomer actuators are a good choice for applying high forces and actuating at high speeds⁷ within low-density and highly stretchable systems. To transport fluid for actuation, fluidically pressurized elastomer actuators require fluidic channels and chambers,^8,9 the design and fabrication of which limit their geometry, often necessitating complex molding and/or assembly to form fully 3D machines.

Foam actuators simplify this fabrication process by eliminating the need for designed channels and chambers. The elastomer foam itself serves as the inflation chamber, while the interconnected porous network serves as fluidic channels. When sealed within a nonporous elastomer, these foams form a fluidic actuator enabling 3D soft machines without complex assembly^10,11 or investment casting.^9,12 While these actuators can function using air, water, and other fluids as the pressurizing medium, we used compressed air as the actuating fluid for all actuators due to the inviscid nature of air and fast transport through micropores.¹³ These foam actuators, however, experience reduced actuation amplitudes. Because they comprise ≥60% void space, less material participates in stretching, making the foams prone to tearing at lower inflation pressures.

In this study, we have improved upon our previous work on pneumatically powered foams^13–15 by buckling the foams before actuation. When the foam is compressed in the direction opposite to our intended actuation mode, the foam attains a higher apparent elongation before break with an added benefit that the elasticity aids in actuation. While compression of porous elastomeric structures (i.e., the controlled buckling of the foam struts) has been shown to produce negative Poisson's ratio behavior,¹⁶ shape memory properties,^17,18 directed deformation,^19,20 and actuation,^21,22 to the best of our knowledge, this article is the first to employ precompression to improve pneumatic actuation. In this study, we show that mechanical foam compression imparts buckling of the microscale struts leading to increased apparent elongation and augmented actuation (i.e., either increased displacement or reduced actuating fluid pressure).

To mold these actuators, we used an elastomeric, polyurethane (PU) open-celled foam (FlexFoam-It!; Smooth-On, Inc.) to form the inflation chamber. Because PU foams have not previously been used for fluidic actuation, to our knowledge, it is important to consider their method of foam formation as it relates to the final properties of the actuator. The foam's liquid precursors consist of three parts: (i) hydroxyl-functionalized prepolymer, (ii) water, and (iii) aromatic diisocyanate cure agent. The water serves as a chemical blowing agent that when reacting with an isocyanate at standard temperature and pressure (STP), generates carbon dioxide in an exothermic reaction, which expands to form the cells of the foam.²³ The use of elastomeric PU foams makes the fabrication of these actuators (Supplementary Fig. S1; Supplementary Data are available online at www.liebertpub.com/soro) faster than our prior method of dissolving or vaporizing embedded porogen salts.^13–15 In addition to rapid fabrication, the PU foam actuators are functional at a much higher porosity (Φ_pore ∼0.93, prebuckled) than our previously reported polydimethylsiloxane foam actuators (Φ_pore ∼0.70), allowing for rapid actuation (∼86% of maximum strain, ɛ = 0.15 in 320 ms; Supplementary Fig. S2) and lightweight devices (∼0.65 g/cm³).

As a potential application for these buckled foam actuators, we have also developed a patient-specific, direct cardiac compression (DCC) heart assist design. DCC devices are implanted mechanical pumps that increase blood flow in failing hearts by applying epicardial compression. Promising pneumatically actuated DCC device examples include designs with an inflatable plastic envelope within either a rigid shell²⁴ or a shape memory wire scaffold^25,26 surrounding the heart and a fully compliant design that uses an assembly of contracting pneumatic actuators.^27–29 The most recent of these devices demonstrates the strong need for highly complex and compliant shapes; this fluidically powered soft machine demonstrated viability through in vivo testing on a porcine heart.²⁹ Our design is similar in material properties, but it uses an elastomer foam for easy, reliable, and reproducible fabrication. Additionally, the objective of our design is primarily to address the shortcomings of other DCC devices—they largely neglect variations in patient-to-patient heart geometry and apply global compression, potentially adversely compressing coronary arteries.^30,31

In this work, we present a DCC device design with the unique features of patient-specific fit and rapid, efficient local actuation. To form the patient-specific device, we use conventional clinical imaging techniques such as magnetic resonance imaging (MRI) or computed tomography (CT) to capture the geometry of the patient's heart (Fig. 1A). We then use the digital model of the heart as the basis of a 3D printed mold for device fabrication. The device employs pneumatically inflated, soft foam chambers (Fig. 1B, C) to apply mechanical compression on the heart muscle adjacent to the ventricles. We have demonstrated rapid DCC device fabrication (<36 h), which enables the potential for on-demand device manufacturing. Furthermore, the portion of the device that contacts the heart comprises entirely soft compliant materials that exhibit similar mechanical properties to biological tissue.¹

FIG. 1.

The direct cardiac compression device concept. (A) The device fabrication process involves the following steps: collection of a chest MRI or CT of the patient's heart, reconstructing a digital 3D model of the heart, digitally isolating the heart, 3D printing a mold from the digital heart model, casting elastomer foam within the mold, and cropping and assembling two foam chambers to form the final DCC device. Scale bar, 3 cm. (B) A device schematic showing elastomer and strain limiting layers surrounding the foam actuation chambers. Scale bar, 250 μm. (C) The device before inflation (left) and during inflation (right) with arrows showing volume displacement upon pressurization. Scale bar, 3 cm. CT, computed tomography; DCC, direct cardiac compression; MRI, magnetic resonance imaging. Color images available online at www.liebertpub.com/soro

By incorporating buckled foam actuators within the DCC device, it is capable of displacing large volumes of liquid (∼70 mL) when inflated while remaining thin (device thickness, L ∼15 mm) when deflated. Additionally, while deflating, the foam passively expels air due to its inherent elasticity. For these reasons, elastomer foams have potential operational advantages over the inflatable plastic envelopes that have been used in other designs.^24–26

Although there are existing long-term, implantable (i.e., biocompatible) PU elastomers,³² we chose a commercially available PU foam for this study because it rapidly fills mold cavities, cures in less than 1 h, produces a low-modulus open-celled foam with high porosity, and is of low cost. Although the materials we used in this study have not yet been tested for chronic tissue compatibility, they serve well as initial surrogate materials, in that their mechanical properties remain stable over the range of temperatures and strain rates required in the physiological environment (Supplementary Fig. S3).

Materials and Methods

Material preparation

We used all foams and elastomers as received and according to the manufacturer's instructions. We mixed the foam precursors vigorously by hand for 30 s before casting. All foam samples were FlexFoam-iT! III (Smooth-On, Inc.), except for the extending actuators which were FlexFoam-iT! V (Smooth-On, Inc.). We used a commercially available PU elastomer (VytaFlex 20; Smooth-On, Inc.) to form the nonporous impermeable coating around the foam. We selected this elastomer due to its high extensibility, ɛ_ult =  ∼10, and tear strength, Γ = 10.5 kN/m. We mixed the PU elastomer parts A and B in a Thinky™ planetary centrifugal mixer at 2000 rpm for 30 s before casting onto the foam. For extending actuators, we added a flexibilizing agent (So-flex, Smooth-On, Inc.) to the sealing elastomer precursor in a 1:1 weight ratio with VytaFlex part B to lower the elastomer modulus.

Actuator fabrication

We constrained the foam in compression by applying a thin coat of PU elastomer to the exterior of the cast foam before applying mechanical compression and holding until the elastomer had cured. To form the final hermetic seal, we cast a second elastomer coating (thickness, t ∼2 mm) around the entirety of the foam. In the case of the extending actuators shown in Figure 3, we also wrapped aramid thread (KEV138NATL01B, The Thread Exchange) circumferentially around the actuator to limit radial expansion upon inflation. Additionally, we designed all actuators to actuate parallel to the foam's rise direction. Further details of this fabrication process are in the Supplementary Data.

FIG. 3.

The effect of foam compression on actuation. Extending actuation of an uncompressed (left) and compressed foam actuator (right). Scale bar, 5 mm. Color images available online at www.liebertpub.com/soro

Tensile testing

We performed all tensile tests according to ASTM D412 on a Zwick Roell z010 instrument pulling parallel to the foam's rise direction. All tests used a 10 kN load cell at a strain rate of 5/min.

Statistical analysis

Statistics presented are mean ± standard deviation with N = 3 unless otherwise noted; p values were based on two-tailed Student's t-test with assumed unequal variance within the sample sets.

Microcomputed tomography imaging

We used X-ray microcomputed tomography (μCT) (ZEISS Xradia Versa 520) to image the internal structure of the foams. We used a previously described MATLAB program¹³ to threshold the μCT image stacks and used ImageJ image analysis package with the BoneJ plugin³³ to determine pore characteristics. Details of the image analysis are provided in the Supplementary Data.

Blocked force testing

We performed blocked force tests by first measuring the maximum extension at a given air pressure. We then depressurized the actuator, set it on a balance with the top surface blocked with a weighted plate to ensure zero extension, and subsequently repressurized it while measuring the force exerted.

Actuator modeling

The 1D actuator model was implemented in MATLAB. We used a large deformation formulation with total stretch on the material taken as the product of the actuation stretch and precompression as relevant. We extracted foam and coating mechanical properties from their respective uniaxial tensile tests. The 1D model approximated the full 3D behavior reasonably well because the Kevlar thread provided most of the radial expansion resistance and foams have a relatively small Poisson's ratio in compression.

DCC device fabrication

Supplementary Data contain details of the conversion of an MRI or CT scan into a digital model. From the digital model of the heart, using Autodesk Maya^® software and an Object30 (Stratasys) 3D printer, we designed and fabricated two sets of molds: one mold designed for casting a 20-mm-thick foam shell surrounding the heart geometry and one for compressing that foam to a thickness of 10 mm. The inset portion of the mold (i.e., the geometry of the heart) was the same for both mold sets. We first cast the PU foam into the 20-mm-thick mold. From that foam, we cut out two sections that surrounded the exterior of the ventricles and avoided the coronary artery regions. We then coated the foam sections in PU elastomer and compressed them into the 10-mm-thick mold for the duration of its cure. After demolding, we painted on an additional layer of PU elastomer (final thickness ∼2 mm) and manually applied a coat of 5% (by wt.) chopped carbon fiber (Fibre Glast) in PU elastomer to the external surface so that all inflation would be directed inward.

Results

Compressed foam actuation

The void space within the foam, in addition to providing an inflatable porous network, can be exploited for mechanical advantage (Supplementary Fig. S4). To achieve this gain, we constrain the foam in a state of mechanical compression using a PU elastomer. The practical limit of uniaxial compression (i.e., complete pore collapse) is the densification strain , which we found to be  = −0.82 and  = −0.70 for foams having a density of ρ = 0.05 and 0.08 g/cm³, respectively (Fig. 2A). As we require an open porous network for inflation, we operated below for all samples.

FIG. 2.

The compressive behavior of polyurethane foams. (A) Mechanical compression of the foams used in this study, (B) a schematic of constrained compression method, and (C) a foam cube before (left) and after constrained triaxial compression to ∼40% original volume (right). Scale bar, 5 mm. (D, E) μCT 3D reconstruction (left), a representative 2D slice (right), and pore size distribution (lower) of uncompressed and triaxially compressed foam. Scale bars, 500 μm. μCT, microcomputed tomography. Color images available online at www.liebertpub.com/soro

By constraining the foam in a compressed state (Fig. 2B, C), we observed an increase in the structure's apparent ultimate strain. Specifically, for foam samples constrained uniaxially to a compression ratio (R_c = ) of 1.5, we observed a significant (p = 0.03, N = 5) increase in apparent ultimate strain ( to ) in uniaxial tension. This difference is nearly equal to the strain increase we would expect from the reversible unbuckling of this compression as shown in Equation (1) below:

We attribute the discrepancy between the measured strain increase ( and ɛ_expected = 0.5 to a slight macroscopic creasing of the compressed tensile samples that initiates tearing at a lower strain (Supplementary Fig. S5). The improved elongation appears to be from a reversible structural change (i.e., strut buckling), supported by the lack of a significant difference (p = 0.39, N = 5) between the ultimate tensile stress of the samples:  = 0.134 ± 0.024 MPa and 0.148 ± 0.024 MPa.

Figure 2D and E shows a comparison of the porous structure of the uncompressed (i.e., R_c = 1) and compressed foam (R_c = 2.5) obtained from a μCT scan. Consistent with pore collapse, we noted a decrease in the average pore size of the compressed foam (mean pore diameter,  = 98 ± 65 μm) compared with the uncompressed foam ( = 165 ± 144 μm); however, the compressed pore network remained highly porous ( = 0.85, measured via μCT stack image analysis), enabling ample open pathways for inflation. Due to the foam's high initial porosity ( = 0.93), we expect an inflatable open-celled network even after large volumetric compression. For example, we estimate that a foam compressed to 20% of its initial volume (i.e., R_c = 5) would have a porosity Φ ∼0.65 (details in Supplementary Data).

The actuators can gain a mechanical advantage by starting with the foams in a compressed state. To demonstrate the more efficient actuation of actuators fabricated from compressed foams, we fabricated two extending actuators (Fig. 3); one with R_c = 1 (no compression) and one with R_c = 1.5 (with compression applied uniaxially). When inflated to equal pressures, the buckled foam (R_c = 1.5) actuator showed much larger displacement than the R_c = 1 actuator. We quantified this behavior through a blocked force test measuring both the maximum displacement (under zero load) and the maximum applied force (when constrained to zero displacement) for each actuator as a function of inflation pressure. Representative data from these measurements are shown in Table 1 with full results shown in Supplementary Fig. S6; Supplementary Table S1. These results indicate that for a given inflation pressure, the compressed foam actuator exhibited ∼2 times the displacement and ∼1.4 times the applied force as the one without compression. We measured this increased efficiency over uncompressed actuators in terms of displacement or force for all tested pressures in the range of 10–90 kPa.

Table 1.

Experimental Measurements and Model Predictions for Maximum Force and Actuated Strain of Uncompressed (R_c = 1) and Compressed (R_c = 1.5) Actuators

	Rc = 1		Rc = 1.5
Inflation pressure (kPa)	Maximum applied force (N)	Maximum actuated strain	Maximum applied force (N)	Maximum actuated strain
Experiment
10	1.3 ± 0.4	0.017 ± 0.004	2.2 ± 0.3	0.038 ± 0.010
50	11.8 ± 1.8	0.082 ± 0.009	16.5 ± 0.2	0.182 ± 0.019
90	25.5 ± 4.4	0.155 ± 0.005	36.5 ± 4.6	0.314 ± 0.033
Model
10	2	0.038	3.14	0.067
50	10	0.184	11.24	0.228
90	18	0.327	19.34	0.391

To further understand the actuated motion, we constructed a 1D statics model to predict actuation strain as a function of inflation pressure and externally applied force. The force balance indicated that the force in the foam, the force in the coating, and the externally applied blocking force must sum to the force provided by the inflation pressure (). The stresses in the foam () and coating () were each related to their strain by the measured uniaxial stress–strain behavior. The foam strain was taken to be the actuation strain minus the precompression strain, so the required stress (and consequently required inflation pressure) was much lower to produce a given actuation strain when the precompression was large. This equation also showed the direct trade-off between blocked force and actuation strain at a given pressure. The model results for uncompressed (R_c = 1) and compressed (R_c = 1.5) foam actuators for the particular materials and geometry used in the experiment (Supplementary Fig. S7) show an expected 20% increase in actuation strain from the precompression. This model also indicates that the larger the radius-to-coating thickness ratio, the greater the effect.

DCC device design

As an excellent example of complex machines that could benefit from soft robotics, we applied our buckled foam system to a cardiac assist device. Specifically, we used the more efficient and thin compressed foam actuators to form a prototype DCC device. This device design applies compression to the ventricular walls, aiding the expulsion of blood. Like other pneumatically driven DCC devices,^24,26 this design avoids direct blood contact, mitigating risks associated with thrombosis and anticoagulation therapies that are present in common ventricular assist device (VAD) designs.^30,34 The use of buckled foam actuators has the potential to provide the following unique features: (i) application of localized compression on the ventricles, avoiding potentially harmful compression of coronary arteries and atria; (ii) the soft compliant materials easily deform in response to anatomical forces; (iii) the low tangent modulus (E < 1 MPa) and highly porous foams (Φ >0.93) allow for rapid inflation and deflation at physiological rates (Supplementary Movie S1); and (iv) they can be molded to fit each patient, enabling physician-directed digital shape modification before surgery to promote reverse remodeling or inhibit ventricular aneurysms.

To demonstrate this potential, we developed a procedure to produce patient-specific 3D printed molds from standard medical imaging scans, including MRI and CT (Supplementary Fig. S8), or 3D optical scans from excised hearts (Fig. 4A), and formed a DCC device (Supplementary Fig. S9). The process began with isolating the heart as a digital model (details in Supplementary Data). From this model, we designed and fabricated 3D printed molds (Fig. 4B) to cast a foam shell around the heart geometry. From the foam shell, we excised two foam sections in regions corresponding to both ventricles while avoiding areas containing coronary arteries. These two foam chambers formed the active expanding (when inflated) portions of the DCC device; the area between the chambers (i.e., the coronary artery regions) is passive elastomer. We then constrained the foam into a smaller mold (R_c = 2) to buckle the foam struts (Fig. 4C). This R_c value was chosen to achieve the needed volumetric displacement based on the ultimate elongation of the foam. Although we did not optimize the R_c parameter, we could tune the parameter to increase or decrease the displacement in future versions of the device. We compressed the foam in a direction radial to the center of the heart and expect the primary expansion of the foam to be in this same direction. We then applied a strain-limiting layer of chopped carbon fiber to the exterior of the foam, which is necessary to direct the expansion of the inflated foam toward the device interior. We note that since the carbon fiber forms a transverse isotropic layer, the device applies only radial compression and does not twist (as the heart does during contraction). Finally, we hermetically seal the device by applying an outer coat of PU elastomer (Fig. 4D, E).

FIG. 4.

Fabrication of the DCC device. (A) The porcine heart that served as the basis for the digitally designed mold (B) that was used to form an organ-specific foam shell. Scale bar, 2 cm. (C) The compressed foam inflation chambers that were coated in carbon fiber and sealed with elastomer to form the final DCC device (D) that fit well around the original heart (E). Color images available online at www.liebertpub.com/soro

We inflated the DCC device using pressurized house air run through a regulator (model 900 dispenser, EFD-Nordson). We controlled the airflow using a three-way valve (04F30C2104AF4C05; Parker), which was activated by a microcontroller (Arduino Uno) through a relay (SRD-05VDC-SL-C; Songle). The microcontroller enabled us to dictate the frequency and duration of pulsatile inflation to mimic physiological rates. We oriented the three-way valve within the air line such that it allowed pressurized flow (for inflation) when activated and allowed venting to the atmosphere (for deflation) when deactivated. An in-line pressure sensor (PSE530-R06; SMC pneumatics) provided the inflation profile within the foam chambers.

We controlled inflation rhythms through either microcontroller code or real-time electrocardiographic (ECG) sensing (Supplementary Movie S1). While microcontroller coding allowed a stable, static, pulsatile inflation frequency and duration for benchtop tests, real-time ECG sensing is more useful within the scope of DCC as a therapeutic tool. When applying compression to the exterior of a beating heart, it is key for the device to compress in phase with the heart's rhythm.

To measure the ECG signals of the heart, we used a prepackaged heart rate monitor (AD8232; SparkFun) with electrical leads arranged in Einthoven's triangle (i.e., a constantly negative lead on the right wrist, a switching polarity lead on the left wrist, and a constantly positive lead on the left ankle). This monitor collected an ECG signal that we used to control the solenoid valve, which triggered inflation of the device.

To demonstrate this DCC device, we designed one to fit a porcine heart (Supplementary Fig. S9) for an ex vivo test. We aimed to mimic the hemodynamics of a young pig, including a heart rate of 90 bpm, stroke volume of 70 mL, and systole and diastole durations of 220 and 440 ms, respectively.³⁵ Using water displacement to measure the inflated foam volume, we observed ∼70 mL displacement for each chamber within a 200-ms time frame (Supplementary Fig. S10). When operating the device outside of the water, we observed (via the pressure profile) that the chambers inflated and deflated rapidly at speeds sufficient to mimic a porcine (or human) heart rate. We coordinated the inflation and deflation rhythm with a simulated electrocardiogram showing sufficiently rapid inflation to align with a native systole and diastole, respectively (Fig. 5A–C).

FIG. 5.

DCC device benchtop and ex vivo performance. (A) Photo sequence of chamber inflation and deflation corresponding to maximum and minimum actuation pressures (B) recorded by an in-line analog sensor. The shaded areas indicate the duration of inflation triggered by a simulated ECG reading (C). Scale bar, 2 cm. Video available in Supplementary Data. Color images available online at www.liebertpub.com/soro

Finally, we performed an ex vivo study using the porcine heart that served as the model for the device. This experimental setup (Supplementary Fig. S11) included a pressure sensor (model ITV1031-21N2BL4, SMC) on the pneumatic line to the device and a flow sensor (SONOFLOW CO.55/120; Sonotec) on the pressurized water lines integrated into and out of the left ventricle, all of which provided real-time measurement of the air and blood surrogate (water) flow. The device's elasticity secured the heart; however, after observing the heart slightly shifting during pressurization cycles, we added a string harness to inhibit this motion.

Within the ex vivo study, manual massage of the heart produced ∼0.18 L/min peak flow, while the DCC device produced ∼0.12 L/min peak flow (Fig. 6; Supplementary Movie S1) at a frequency of 60 beats/min. Although these flows are significantly lower than the native porcine cardiac output (∼5 L/min), this experiment demonstrated that the device is able to produce a repeatable, consistent flow profile and is able to impart ∼67% the flow of manual massage. While the measured flow rates through the heart were lower than expected, some decrease in performance is attributed to the decreased compliance of the cadaveric heart relative to its living properties. An in vivo study, therefore, may show increased performance. We also note that the DCC device we fabricated was molded from a contracted heart; this choice ultimately limits the flow rate of our pump. As this article is focused on the application of buckled foam actuators and not on device efficacy, future work will focus specifically on improving the DCC device design.

FIG. 6.

Ex vivo demonstration on porcine heart. (A) The device setup for the ex vivo demonstration on the porcine heart with (B) the collected air line pressure measurements and corresponding water flow rate during ex vivo demonstration on the porcine heart. Video available in Supplementary Data. Color images available online at www.liebertpub.com/soro

Conclusions

Buckled foam actuators exhibit improved force or displacement behavior over previous foam systems; they are more efficient because they either provide a higher force or displacement at a given inflation pressure or require a lower pressure for a given force or displacement. Although we demonstrated this effect using PU foam and elastomer, the fabrication process is general and should be compatible with any variety of other open-celled elastomer materials and fluidic actuator designs.

We attribute the increased actuation behavior to the foam's elasticity. Constraining the foam in compression generated residual stresses that provided a restoring force toward the original unstressed state. Because we compressed the foam in the direction opposite to actuation, the returning force is in the direction of actuation, aiding in this motion. We note that the returning force will only be present up to the state when the foam is fully unbuckled. Another benefit of buckled foam actuators is that the imparted residual stresses limit the effect of foam strain hardening. Because the compressed foam is under a lower strain when actuated, the foam tangent modulus is also lower and provides reduced resistance to expansion upon inflation.

Our demonstration DCC device formed a close fit around the porcine heart. In an ex vivo experiment pumping fluid through the heart, the device achieved ∼67% of the peak flow rate obtained through manual massage of porcine ventricles at a duration of 500 ms for systole and 500 ms for diastole. Although we designed the device to fit a contracted heart, investigating device geometries based on different stages of the cardiac cycle would be an important direction for future studies. For example, designing the device from a heart at the end of diastole may not only minimize device effects on ventricle refilling but may also pose a risk of forming a less conformal fit around the heart.

Fabricating the device to be patient specific enables not only an optimal fit within the patient's chest cavity but is also a step toward two future possibilities. First, because this device is produced from a digital model of the patient's heart, it is possible for a physician to digitally modify the final geometry before device fabrication. This would be particularly important for patients exhibiting dilated cardiomyopathy (an enlargement of the heart, reducing its pumping efficiency). A patient-specific DCC device could be formed to the heart's healthier nonenlarged shape, potentially promoting reverse remodeling of the cardiac tissue.³⁶ Second, this fabrication method does not limit the device size allowing for potential adult and pediatric use alike; furthermore, for pediatric care, its compliant properties could potentially be harnessed to grow with the patient.

Additionally, the high porosity of these open-celled PU foams allowed the rapid inflation and deflation that are critical for mimicking cardiac dynamics. The repeatable ∼70 mL inflation displacement within a 200-ms time frame closely mimics the heart's systole in terms of stroke volume and duration.

The fabrication time (<36 h) is currently limited by the process of 3D printing of the molds. We believe this time could be reduced to <24 h by implementing new, rapid 3D printing techniques that dramatically reduce printing times.³⁷

A limitation of this DCC demonstration is that the biocompatibility of its components, to our knowledge, is not yet known. Future studies should integrate known compatible materials with accompanying fatigue tests for measuring durability over hundreds of millions of cycles (a useful lifetime for a VAD). Additionally, we intend to perform the device demonstration using fresh porcine hearts in the future in an attempt to better approximate the mechanics and dynamics of living tissue.

Acknowledgments

The authors thank Andrea Biondi and Joe Clifford for assisting in developing the MRI-digital model process, Teresa Porri for performing the μCT imaging, Roger De la Torre for aiding with the ex vivo test setup, and Jonathan Butcher and Anthony Ryan for their valuable discussions. This work was supported, in part, by the Air Force Office of Scientific Research under award number FA9550-15-1-0160, by a 3M Non-Tenured Faculty Award, and a Weill Cornell Medicine CTSC TL1 traineeship (NIH/NCATS grant no. TL1TR000459). This work made use of the Cornell Center for Materials Research Shared Facilities, supported through the NSF MRSEC program (DMR-1120296). Imaging data were acquired in the Cornell BRC-Imaging Facility using the shared, NIH-funded (S10OD012287) ZEISS Xradia Versa 520 μCT.

Author Disclosure Statement

No competing financial interests exist.