A novel soft cardiac assist device based on a dielectric elastomer augmented aorta: An in vivo study

Abstract Although heart transplant is the preferred solution for patients suffering from heart failures, cardiac assist devices remain key substitute therapies. Among them, aortic augmentation using dielectric elastomer actuators (DEAs) might be an alternative technological application for the future. The electrically driven actuator does not require bulky pneumatic elements (such as conventional intra‐aortic balloon pumps) and conforms tightly to the aorta thanks to the manufacturing method presented here. In this study, the proposed DEA‐based device replaces a section of the aorta and acts as a counterpulsation device. The feasibility and validation of in vivo implantation of the device into the descending aorta in a porcine model, and the level of support provided to the heart are investigated. Additionally, the influence of the activation profile and delay compared to the start of systole is studied. We demonstrate that an activation of the DEA just before the start of systole (30 ms at 100 bpm) and deactivation just after the start of diastole (0–30 ms) leads to an optimal assistance of the heart with a maximum energy provided by the DEA. The end‐diastolic and left ventricular pressures were lowered by up to 5% and 1%, respectively, compared to baseline. The early diastolic pressure was augmented in average by up to 2%.


| INTRODUCTION
Congestive heart failure (CHF) is a progressive and debilitating condition affecting a substantial proportion of the elderly population. It is estimated that CHF causes more deaths than all cancer cases combined, and that approximately 26 million people worldwide suffer from the disease. A total of 5.8 million of these patients reside in the United States, with an estimated economic impact of 40 billion dollars on the country's economy in terms of medical costs and productivity loss. 1 While drugs are used to improve heart function and relieve symptoms, they usually fail to restore cardiac function in the long term. Heart transplantation is the gold standard for patients with severely impaired left ventricular dysfunction who cannot be treated otherwise. Due to the lack of donor organs, however, cardiac assist devices for permanent hemodynamic support remain an unmet clinical need. A safe and fully implantable system, capable of restoring cardiac function and thus eliminating the need for heart transplantation, should be used to achieve this objective. 1 Currently, existing cardiac assist devices are used as bridge to recovery, bridge to next decision, bridge to transplantation, and, more recently, as destination strategy. 2 Among these devices, ventricular assist devices (VADs) based on rotary elements (axial and centrifugal pumps) are currently the most widely adopted solutions. Although pulsatile flow may be preferred over continuous flow, as it is closer to normal physiology and favors optimal tissue perfusion, 3 current VADs can only ensure a constant laminar flow. 4 Moreover, due to high shear stress on the blood cells caused by the rotating parts, they are associated with a higher risk of hemolysis and thrombosis, which forces patients to depend on permanent oral anticoagulation for the duration of the support. 5 In this feasibility study, we therefore focus on the aortic counterpulsation (ACP) principle, which does not require anticoagulation therapy in the majority of cases 6 and help to preserve or even enhance pulsatile flow in the aorta.
The basic principle is similar for all ACP devices: (i) they augment early diastolic pressure to increase flow in the coronary arteries and (ii) they reduce the resistance to systolic output (afterload). Depending on their location within the aorta, ACP devices are classified as: intraaortic, 7,8 extra-aortic, [9][10][11] and para-aortic 12 counterpulsators. The different characteristics of these devices are summarized in Table 1.
Among all the ACP devices, the intra-aortic balloon pump (IABP) has been the most widely used in clinical settings since decades. Its balloon is inflated in the aorta during diastole, which leads to augmentation of the diastolic pressure, while the presystolic deflation of the balloon leads to a sudden reduction of the afterload immediately before the opening of the aortic valve. 7,8 In general, IABP are used in high-risk patients as a short-term support to other solutions (e.g., VADs). However, patients may suffer from bleeding, hemorrhage, infections, limb and mesenteric ischemia. 6 To overcome these significant limitations, extra-aortic counterpulsation devices were designed by wrapping and compressing the aorta externally during diastole. In this direction, C-pulse is an extra-aortic balloon (EAB) counterpulsation device, pneumatically actuated, which has shown promising results 9 in terms of improved coronary flow and ejection fraction. However, in the context of long-term cardiac support, the pneumatic actuation of existing ACP devices is a limiting factor due to the large size of the gas sources. Large transcutaneous pneumatic drivelines (greater in size than the electric lines used for VADs) are required and can cause severe infections. Additionally, pneumatic actuation leads to long inflation times and can introduce uncontrollable delays in the device operation. Thus, a fully implantable solution for these devices is difficult to envision.
Dielectric elastomer actuators (DEAs) have emerged as a noteworthy alternative solution to build cardiac assist devices because they are soft devices (in contrast to rigid VADs) and only require an electric signal to be actuated. DEAs consist of a hyperelastic dielectric membrane placed between two compliant electrodes. When a voltage is applied between the electrodes a pressure known as a Maxwell pressure compresses the dielectric elastomers in the thickness direction allowing it to expand in the other directions. 13 In our previous study, 14 we introduced an augmented aorta based on such DEAs.
When exposed to the blood pressure, the soft device passively inflates. The activation of the device allows to expand the device further and thus to control the compliance of the system. DEAs' low weight and the absence of bulky drive lines are promising advantages compared to existing technology in view of a fully implantable system and long-term cardiac support. Moreover, as only electrical activation is needed, the activation time of the device can be much lower than for pneumatic actuators and trans-cutaneous wireless power transfer could be achieved 15 to reduce infections.
In this study, we tested the feasibility of the DEA augmented aorta in an acute in vivo porcine model. The DEA was implanted in the descending aorta of five pigs. The aim was to show that the DEA could provide support to the heart by lowering the end-diastolic pressure, and augment coronary flow by increasing early aortic diastolic pressure. Additionally, the optimal activation and deactivation parameters in terms of activation duration and delay relative to aortic valve opening were investigated and reported.

| The DEA acts as a counterpulsation system
The DEA is a cylindrical tube that replaces a section of the descending aorta to support the heart by counterpulsation. In its passive state, the compliant nature of the DEA mimics the elastic behavior of the aorta, storing energy during systole and releasing it during diastole. Figure 1a shows the activation timing of the DEA when operating as a counterpulsation device as described in the literature. 14  This activation scheme has two main advantages. Figure 1b shows the pressure-volume characteristics of the DEA compared to the aortic characteristics. level, the device can store more blood and decrease aortic pressure by activating it just before systole. By releasing the additional blood volume stored in the DEA at the beginning of diastole, the pressure in the aorta increases, thereby, augmenting the coronary flow.

| Dielectric elastomer augmented aorta
For the DEA to work as an augmented aorta, two issues must be addressed: maximizing the actuation effect of the device and ensuring good sealing of the device anastomosis with the aorta. To obtain high displacement while keeping the activation voltage as low as possible, the DEA was manufactured as a multilayered device. 16 The DEA comprises a stack of four Elastosil films from Wacker Chemie The biggest diameter of the connector, close to the device, matched the internal radius of the DEA ( Figure S3). The smallest diameter section was designed to fit into the aortic lumen. During the in vivo experiment, this section of the connector was inserted into the aorta and a plastic clamp tie was tightened around the aorta for the anastomosis (Figure 1e). The small step at this end of the connector ensured the sealing of the anastomosis. The external housing limited the radial expansion of the DEA to avoid electromechanical instability that could lead to early breakdown of the actuator. 18 Figure 1d shows the overall dimensions of the device.
Two different types of DEAs were made with 29 mm or 39 mm active electrode lengths corresponding, respectively, to 40 and 50 mm total device lengths. The overall thickness and thus the compliance of the DEA were designed to be rigid enough when passive while still deforming sufficiently when actuated. In the following, we will only refer to the DEA by their active lengths. In both cases, with the selected designs and the blood pressure levels, the DEA requires very high-voltage levels (>6 kV) to obtain significant deformation. During the in vivo experiments, the actuation voltage was selected as trade-off between maximum deformation and electric breakdown limit of the device. 18 The first activation level for the measurements started at 6 kV, and it was then increased to 6.5 and 7 kV if the DEA did not break before.

| The device is synchronized to the heart cycle in vivo
We tested the DEA in the descending aorta in a porcine model (n = 5). The pigs (Edelschwein pigs, 50 kg, mixed sex) had the DEA implanted via left-sided thoracotomy (Figure 2a), and the heart rate was controlled using a pacemaker at 100 bpm. The control and data acquisition setup in the operating room enabled recording the left ventricular pressure and volume, aortic pressures and flows upstream and downstream of the device (Figure 2b,c). The DEA was synchronized to the heart cycle using the pacemaker signal as trigger ( Figure 2c) and delayed with respect to the pacemaker signal to synchronize with different time points of the heart cycle. Figure 3a shows pressure, flow, and the pacemaker signal for two consecutive heart F I G U R E 3 Legend on next page. The start and end times were tested before (À5% of heart cycle), at (0% of heart cycle) and after (5% of heart cycle) aortic valve opening and closure, respectively, leading to a variable actuation duration (all combinations are listed in the table of Figure 3c). Before and after each device actuation period (for both protocols) the device was turned off, and these heart cycles were used as baseline for each actuation period. Figure Figure S12) and was not used in the analysis. Figure 4a shows the averaged heart cycles, and how the parameters EDP and AoP dia are defined.

| Targeting the best actuation scheme
The actuation of the DEA for different phase shifts (Protocol 1) changed the pressures and flow in the aorta differently compared to baseline. Figure  In this protocol, the actuation was mainly during systole and start and end times of DEA actuation were delayed with respect to aortic valve opening and closure, respectively, with À5% (before), 0% (at) and 5% (after) delay in percentage of heart cycle; 0% delay corresponds to synchronization with aortic valve opening and closure for the start and end times, respectively (considering the pressure delay in the DEA, with earlier activation to allow the pressure wave time to reach the AV at opening). The two protocols were performed in all animals as seen in the overview. aggregated). compared to baseline for the phase shifts 0%-45% ( Figure 4b and  The device also affected the early diastolic aortic pressure (AoP dia, Figure 4b) depending on the phase shift. An augmentation of the early diastolic pressure of 0.2%-2% compared to baseline was seen for phase shifts with the end of actuation during late systole or early diastole (80%-15%, p < 0.05). An augmented aortic pressure during early diastole is expected to increase the coronary flow. For the average aortic flow (avg. Qao), it can be seen in Figure 4b that there is a small increase or decrease of the aortic flow for some cases (e.g., 15% and 40% phase shift). However, when considering the results for all the DEAs there were large variations. For 5%-10%, there was a small significant increase in flow compared to baseline ( average by 5%-7% (at 90%-0% phase shift), the LVP was lowered by 1%-2% (at 0%-10% phase shift), and the early diastolic pressure was increased by 1%-2% (at 0%-10% phase shift). Figure S11 shows the raw pressure and flow signals recorded in animal 4, DEA 2 and the clear impact of the DEA actuation.

| Interplay between the device and the cardiovascular system
In addition to the evolution of physiological parameters, we evaluated the influence of the DEA by estimating the energy it provided to the cardiovascular system. Figure 1b Figure S4B shows the picture of the setup. Figure 5a shows the pressure-volume characteristics of 39 mm DEAs for voltages up to 7 kV ( Figure S5 shows The results show that if the DEA was activated when the aortic pressure was low (typically at the end of diastole) and deactivated when the pressure was high (end of systole, early diastole) (e.g., 0% phase shift) the energy provided by the DEA was maximized. For this case, the pressure would increase while the DEA was active leading to larger expansion, and the DEA would release and provide a larger amount of stored energy to the cardiovascular system when deactivated (Figure 5a, green area). This behavior corresponds to a counter-clockwise pressurevolume cycle. However, if the DEA activations started at a high pressure and ended at a low pressure (e.g., 60% phase shift) the DEA was not providing energy but used energy from the cardiovascular system. The pressure would decrease during actuation reducing the expansion of the DEA (lowering the amount of stored energy) and the DEA would work against the heart when deactivated just before the start of systole (increasing EDP) and thereby be detrimental to the assistance (Figure 5a, orange area). This behavior corresponds to clockwise pressure-volume cycle. In Figure 5b, we see the estimations of the energy provided by the different DEAs for the different phase shifts during Protocol 1. For phase shifts between 90% and 40%, the DEA provided energy to the cardiovascular system with maximum values for 90%-20% phase shifts.
These results confirm the timing of activation to maximize the effect of the actuator presented in the previous section.
F I G U R E 5 In vitro estimation of the energy supplied by the dielectric elastomer actuator (DEA) in vivo.
(a) Pressure levels in the DEA during in vivo experiment and the pressurevolume characteristics with the corresponding energy cycle for 0% and 60% phase shift for DEA 2 (39 mm) at 7 kV. The light blue and purple lines represent the pressure levels in the DEA at start and end of activation and deactivation, respectively. The pressurevolume characteristics from 0 to 7 kV were obtained through in vitro measurements prior to the in vivo experiments. Depending on the activation time, the DEA will provide to (green counterclockwise energy cycle) or take (orange clockwise energy cycle) energy from the cardiovascular system. (b) Evolution of the estimated energy provided by the DEA as a function of the shift of activation for all DEAs tested for Protocol 1. The maximum energy is reached for 90%-20% phase shift.
The best start and end times of activation, and thereby the duration of actuation, were investigated by testing three different start and end times (before [À5%], at [0%] or after [5%] aortic valve opening and closure, respectively). Figure 6a shows the averaged pressure and flow waveforms for two different start and end times of activation compared to baseline. The impact of the different start and end times were slightly different due to the timing of activation. Figure 6b shows the average values of the parameters for each DEA (this evolution is further detailed in Figure S10B where all tests are aggregated).

| DISCUSSION
We successfully established a soft cardiac assist device based on a DEA, which replaces a segment of the aorta and works as an augmented aorta. The DEA was implanted and tested in vivo in the descending aorta of an acute porcine model. The developed manufacturing process allowed us to create the DEA as a multilayer rolled tube with an internal diameter comparable to the porcine aortic diameter. The tube was combined with a 3D-printed external housing, which ensured safe operation regarding electromechanical instabilities. Specific connectors have also been designed for the tight anastomosis between aorta and device.
When activated the DEA increases its internal diameter and thereby creates a decompression wave decreasing the aortic pressure.
When deactivated, it returns to its passive state creating a compression wave increasing the aortic pressure. Existing counterpulsation devices (e.g., IABPs or EAB 10 ) obstruct the aortic lumen when activated and this restricts their activation to the diastolic phase only (because if a balloon is inflated during systole, it would increase the afterload due to obstruction) as shown in Table 1. On the contrary, our DEA device allows cardiac assistance without aortic obstruction because the DEA's passive internal diameter is comparable to the aortic diameter. One of the major differences compared to IABP is that, while the classical IABP lies within the aorta, the DEA replaces a short segment of the aorta. In this context, the principle is similar to the para-aortic counterpulsation device, also called para-aortic blood pump (PABP) as both devices replace a segment of the aorta and increase its volume, 12 allowing more freedom in terms of activation and deactivation phases. However, PABP requires a bulky pneumatic chamber hardly implantable in the thoracic cavity, and it can modify greatly the flow of blood from its natural path. Our DEA only needs an electric activation to assist the heart and does not require bulky pneumatic elements as other counterpulsation devices do. Similar attempts were conducted by Hashem et al. 19 and Starck et al. 20 who developed extra-aortic counter-pulsation devices based on: (i) shape memory alloy fibers, which contracts following Joule heating, and (ii) a ferromagnetic silicone that contracts triggered by an external magnetic field, respectively. However, the low compatibility of the working principle with human implantation and the effectiveness of these latter devices need further investigations 21 before in vivo implementation, especially for blood pressures higher than 50 mmHg. 20 In this in vivo study, the nonobstructive nature of the DEA made it possible to investigate the effect of actuation at different time points during the heart cycle (phase shifts). Depending on the actuation phase shift, the level of assistance was changing: with actuation at 90%-10% (counterpulsation) of the heart cycle (0% = actuationstart synchronized to aortic valve opening) the estimated energy provided by the DEA was maximized (Figure 5b) and thereby also the unloading of the cardiovascular system. With actuation start at 60% phase shift (co-pulsation), the estimated energy was minimized and the DEA actually increased the load on the cardiovascular system.
Thus, by controlling the phase shift of actuation, it is possible to control the loading of the heart. Gradual loading of the left ventricle might be a method to stimulate recovery of patients with heart failure. 22 Looking at the measured pressure and flow parameters, the phase shifts 90%-10% seem to be the best to lower EDP and augment the early diastolic aortic pressure (Figure 4b). To optimize the DEA assistance further, different start and end times of activation with the DEA mainly active during systole (comparable to 90%-10% phase shifts) were investigated. A start of activation just before the aortic valve opens (À5% of heart cycle [0% is defined as aortic valve opening]) and an end of activation at or just after aortic valve closure (0% or 5% of heart cycle [0% is defined as aortic valve closure]) were found to be optimal (Figure 6b). In this way, the DEA is activated at low pressure at the end of diastole and thereby decreases the EDP. It is active during systole and thereby lowers the average LVP. It is deactivated at high pressure at start of diastole and thereby increases the early diastolic aortic pressure. Hence, the optimal activation of the DEA helped reducing the left ventricle pressure during systole and the enddiastolic aortic pressure by up to 1% and 5%, respectively, and augmenting the pressure in the aorta during diastole by up to 2%, which might enhance the coronary flow (see Table S2 for maximum, minimum, and average values for all measured hemodynamic variables).
As comparison, using IABP in patients, Kolywa et al. 23  Nevertheless, we were able to show that the DEA can assist the cardiac system in two animals despite these low values. The devices were originally designed for implantation in humans and therefore to work at higher pressures (between 80 and 120 mmHg). At higher pressures, the displacement would have been bigger and thus also the energy provided by the DEA and its impact on the cardiovascular system. In this study, we were limited by the pressure levels in the pigs and the breakdown voltage of the DEA and were not able to reach voltages higher than 7 kV. But we can already see the impact of supplying higher energies. Indeed, we see in Figure 4b that the greatest changes of EDP and early diastolic aortic pressure were obtained for a 39 mm DEA at 7 kV thus with more volume deformation than other DEA. We see similar trends in Figure 6b where the higher decrease of the end diastolic pressure is obtained for the same DEA at 7 kV.
Nonetheless, the influence of the DEA and its energy is also linked to the specific blood pressure levels (which can vary during the in vivo experiments) and can explain the variations for similar size DEAs tested at identical voltages.
The results presented in this study serve as a proof of concept for the use of DEAs as cardiac assist device. However, some aspects still need to be further optimized before considering a potential clinical application. First, an increase of performance is required to justify the implantation of the device in a patient. The first step in this process is an optimization of the design to conform to the patient's pressure levels. By doing this, we expect to obtain higher deformation of the DEA and thus more provided energy while also increasing the stability of the device. Furthermore, we could anticipate that further in vivo experiments should focus on DEA implantation in the ascending aorta.
The benefits of counterpulsation at the level of the ascending aorta are expected to be more important compared to counterpulsation in descending aorta for several reasons 24 : (i) it ensures a better synchronization to the cardiac cycle due to its proximity to the aortic valve, (ii) counterpulsation efficiency and blood volume displacement are maximized because the pulse propagation in other arteries is minimized, and (iii) the increase in coronary blood flow is also maximized, 25,26 more than with IABP. In the current study, we tested our device in anesthetized, but healthy porcine hearts. The aim of further in vivo experiments will be to include hearts with reduced systolic function such as myocardiac infection or ischemia reperfusion.
From the engineering aspect of the device before it could be con-

| Study design
The aim of this study was to show that our soft DEA could provide some support to the heart by lowering the end-diastolic pressure, and augment coronary flow by increasing early aortic diastolic pressure.
An acute porcine model, atrium-paced at 100 bpm, was used to test our device in vivo in n = 5 pigs. The device was implanted in the descending aorta. Cardiac parameters were recorded during baseline (60 heart cycles before and 60 heart cycles after device-actuated period) and during actuation of the device (60 heart cycles). Forty consecutive heart cycles, 20 baseline and 20 device-actuated, for each device protocol were analyzed using MATLAB (MathWorks, Natick, USA).

| Device synchronization and study protocol
The DEA assist device was activated and deactivated using a highvoltage unit, which was synchronized to the pacemaker of the animal using the pacing signal as trigger. To synchronize the DEA activation to the aortic valve opening and deactivation to the aortic valve closure a few heart cycles were measured. Figure Figure 3b shows the concept of the first protocol highlighting three different phase shifts (0%, 45%, and 90%).
Fifteen different phase shifts were performed from 0% to 90%, with steps of 5% from 0% to 50%, and step of 10% from 50% to 90% (except for Animal 4, in which the protocol was performed with steps of 10% only, and thereby for 10 different phase shifts). In the second protocol (Protocol 2), the expected best phase shift with actuation mainly during systole was used with the actuation start and end times synchronized to the aortic valve opening and closure, respectively (taking the pressure delay into account and activating/deactivation the DEA this delay earlier). The start and end times of activation were shifted with À5%, 0% and 5% (of valve cycle) compared to valve opening and closure, respectively, giving a total of nine different recordings (three start times and three end times). Figure 3c shows the concept of the second protocol highlighting the three different start and end times. The tables in Figure 3b,c give the full overview of the two protocols with 15 and 9 different phase shifts and start and end times, respectively.

| Hardware equipment
The complete hardware setup to convey the experimentation was separated in two parts: activation of the DEA and data acquisition (Figure 2c).
A CompactDAQ (National Instruments, Austin, USA) connected to a computer with LabVIEW (National Instruments) was used to output the activation signal to the high-voltage amplifier (Trek 20/20C; Advanced Energy, Denver, USA). The amplifier was connected to the DEA through wires and a small insulating box that allowed to secure the pig and medical staff from the high voltage. The pacemaker had two identical outputs: one for the pacing of the heart and the other for triggering the DEA activation. The activation profile was defined in LabVIEW as shown in Figure S13 by the period of the signal, the rising and falling time, the size of the plateau, the delay for the activation compared to the pacemaker signal, the voltage value, and the number of cycles. The rising and falling times (50 ms) were reduced as much as possible to be closer to a square signal while limiting the current flowing in the DEA that could lead to deterioration of the performances. As the pacing was very repeatable, the pacemaker triggered the activation only once at the beginning and the synchronization was ensured by the identical value between pacemaker pacing period and defined period in the program. To ensure, the signals were not shifted due to electromagnetic compatibility issues, the trigger signal from the pacemaker was galvanically isolated through a magnetic transformer. In Figure S14, we can see the superposition of all the activation signals compared to the pacing signal. We can see a small drift of maximum 3 ms between the 60 activations, mainly due to a slight error between the pacing period and the activation period that we considered not significant in the analysis of data.
All the signals from the experiment were acquired through two Powerlabs (ADInstruments) of 16 and 8 input ports. The latter one was used to obtain the signals from the pressure-volume catheter while the former one acquired the signals from the pressure catheters, the flow probes as well as the pacemaker trigger signal and the current and voltage monitoring values from the high-voltage amplifier. A computer with LabChart Pro (ADInstruments) was used to visualize the signals in real time and save the data at a sampling frequency of 1 kHz.

| Statistics
Pressure, flow, and volume were recorded continuously using Lab-Chart. All protocols were performed for 60 consecutive heart cycles with actuated DEA, with at least 60 consecutive heart cycles baseline before and after the actuated cycles. Data from the two (n = 2) last animals were used in the analysis, due to different problems occurring during the three first experiments (e.g., early electrical breakdown of the device or early death of the animal). This study followed the 3Rs (Replace, Reduce, Refine) principles 27 in which researchers are obliged to keep animal experiments to a minimum. Because this was an in vivo feasibility study (to test whether our DEAs could support the cardiac system) the low number of animals was justified and followed the 3Rs principle. Table S1 gives an overview of all five animals and all DEAs and protocols performed. No useful data could be recorded in Animals 1 and 3, due to early electrical breakdown of the devices and early death of the animal, respectively. In Animal 2, instable hemodynamic parameters during the recordings made analysis of the results challenging, and they were therefore excluded from the reported analysis. Figure S7A,B shows the results for Protocol 1 and 2, respectively, in Animal 2 with DEA actuation at 6.5 kV for two con-

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.