The heart is one of the most important organs in the vertebrate body. The heart and blood vessels make up the cardiovascular system that circulates blood throughout the body to deliver oxygen and nutrients and remove carbon dioxide and other wastes.[1] The heart muscle is powerful and robust, working relentlessly to maintain the proper functions of the entire body. Heart failure (HF) is an emerging epidemic that affects more than 60 million people worldwide.[2] HF is a condition in which the heart is unable to supply sufficient blood to the body as a normal heart.[3] HF symptoms include shortness of breath, swelling in the legs from fluid buildup, fatigue, and poor exercise tolerance. The most common causes of HF are myocardial infarction (heart attack) and coronary heart disease. Other diseases that may lead to HF include diabetes, high blood pressure, arrhythmias (irregular heart rhythm), and cardiomyopathy (heart muscle disease).[4]
HF is a chronic condition that often requires lifelong treatment to alleviate symptoms and improve quality of life. Treatment of HF usually involves lifestyle changes, cardiac rehabilitation, medications for early-stage HF, and surgical intervention for advanced stages of HF.[5–7] Surgical procedures include coronary artery bypass graft for people with coronary heart disease, catheter ablation to treat abnormal heart rhythms, and heart valve surgery to repair or replace malfunctioning heart valves.[8] In some cases of extremely severe HF, a heart transplant may be recommended[9] but is limited in availability due to donor shortage and many patients who may otherwise benefit remain untreated. Medical robotics has provided a class of implantable devices that support the pumping function of the heart. Some devices are mechanical pumps that take over the heart's function in circulating blood, while others surround the heart to provide additional compression. These devices are designed for the treatment of advanced HF, either as bridging therapy for patients awaiting a heart transplant or as destination therapy. There are broadly three types of mechanical implantable devices for the treatment of HF: ventricular assist devices (VADs); passive ventricular constraint devices (PVCDs); and direct cardiac compression devices (DCCDs).
Ventricular Assist DevicesThe first type, left ventricular assist devices (LVADs), is implanted in the left ventricle (LV) to circulate oxygen-rich blood from the LV to the body through the systemic circulation.[10] Less commonly, VADs are used for the right ventricle (RV) to pump oxygen-poor blood from the RV to the lungs through pulmonary circulation and are called right VADs or RVADs. There are also biventricular assist devices (BiVADs) to assist both ventricles. A VAD typically comprises a mechanical pump, an inflow cannula, an outflow graft, and a drive system. It is usually implanted directly into the ventricles to receive blood through an inflow cannula. The outflow graft returns blood from the pump back to the corresponding blood vessels. The controller system is located outside the body and is connected to the pump via a driveline cable.
Some well-known VADs are Heart Assist 5,[11,12] Jarvik 2000,[13,14] HeartMate 3,[15,16] HeartWare, and TandemHeart. The HeartWare HVAD System (Medtronic, USA) is a mechanical implantable cardiac assist device that uses a centrifugal flow pump implanted directly into the bottom of the LV (the apex). The pump circulates blood from the LV to an outflow cannula connected to the aorta.[17] Several clinical studies have shown HeartWare's capability to provide adequate blood flow, thereby improving symptoms of HF.[18–20] However, Medtronic decided to stop the distribution and sale of the HVAD System in June 2021 due to risks of neurological adverse events, mortality, and potential failure to restart the device.[21] TandemHeart (LivaNova, USA) is an extracorporeal VAD, which means the pump is placed outside the patient's body. The TandemHeart features a centrifugal flow pump to circulate blood, an inflow cannula inserted into the femoral vein, puncturing the septum to access the left atrium, and an outflow cannula inserted into the femoral artery.[22,23] Recently a total artificial heart (TAH) has been developed by BiVACOR Inc, USA, that completely replaces the patient's native heart.[24,25] The BiVACOR TAH is designed for long-term implantation in patients with end-stage HF. The device features a centrifugal pump with a double-sided, magnetically levitated impeller to provide simultaneous perfusion to the systemic and pulmonary circulations.
Treatment of HF with VADs helps alleviate symptoms, enhance heart function, and improve quality of life.[26] However, VADs have hard components that come into direct contact with blood, causing potential adverse events including infection, bleeding, and device thrombosis.[10,27–29] Furthermore, patients are required to take blood thinners during treatment to prevent blood clot formation.
Passive Ventricular Constraint DevicesThe second type of mechanical device for the treatment of HF is PVCD. These devices are constructed from passive elastic materials to limit cardiac enlargement.[30] PVCDs are intended to enclose the epicardium to decrease stress on ventricular walls and reduce heart muscle stretch. PVCDs are recommended for people with LV remodeling and dilated cardiomyopathy (weak and enlarged heart muscle) to prevent further dilation.[31–34]
The HeartNet ventricular constraint device (Paracor Medical, USA) is a flexible band made from nitinol wires coated with silicone. The band is wrapped around the heart circumference to provide pressure to the epicardium.[31] The HeartNet with a high rate of transplant success (98%) has the potential to reverse LV remodeling.[35,36] Acorn CorCap cardiac support device (Acorn Cardiovascular, USA) is a polyester mesh that encloses the entire heart surface from base to apex to limit cardiac dilation.[31] Clinical results have demonstrated the long-term safety of CorCap implantation and the potential for LV reverse remodeling.[37] Instead of making a fixed-size PVCD, Ghanta et al. developed an adjustable ventricular restraint device. The device has a cup shape with two thin layers made from medical-grade polyurethane.[38] The space between two layers can be inflated or deflated to vary the pressure applied to the heart's surface. The restraint level can be adjusted periodically to improve treatment outcomes.
PVCDs surround the epicardium and are not in contact with circulating blood. PVCDs therefore avoid complications of blood clot formation and bleeding. However, the therapeutic efficacy in improving symptoms of HF is limited due to the principle of passive constraint rather than active compression.[39,40] As such, PVCDs are suitable for the treatment of early-stage HF to prevent further ventricular dilatation, thus slowing or ceasing disease progression.
Direct Cardiac Compression DevicesThe third type of mechanical device for HF treatment is DCCDs. These devices typically have conical shapes, allowing easy attachment to the epicardium. Surgical sutures or suction cups are usually required to maintain contact between the device and the heart. DCCDs are designed to actively provide compression to assist the heart during systole to pump blood out of the ventricles.[41] In cases where good adhesion is established between the device and the heart, the device can support diastole to actively fill the ventricles with blood. To synchronize the compression and expansion of the device with the systole and diastole of the heart, an electrocardiogram (ECG) is required in the control system. Treatment of HF with DCCDs eliminates complications from blood contact with the device, which remains a major problem for VADs.[9,29] In addition, DCCDs provide active movements rather than passive constraints like PVCDs, which means that DCCDs have the potential to deliver greater therapeutic effects.[28,42]
The development of DCCDs has a long history, dating back to 1956 with the pneumomassage introduced by Bencini and Parola to massage the heart.[43] Later, the Anstadt cup for cardiac rehabilitation to increase blood pressure and cardiac output appeared in 1965.[44–46] These promising results inspired many subsequent developments. AbioBooster cardiac assist device (AbioMed, USA) was designed to surround the heart for long-term assistance. The device features a number of silicone-coated inflatable polyurethane tubes arranged in a cylindrical shape to enclose the epicardium.[47] When positive pneumatic pressure is applied, these tubes inflate, generating compression to support systole. Conversely, these tubes are deflated by removing the input pressure to accommodate diastole. One study showed that the AbioBooster could produce an aortic pressure of 115 mmHg and a blood flow rate of 6.5 L min−1.[48] In a slightly different approach, HeartPatch (Heart Assist Technologies, Australia) used two independent inflatable silicone patch components attached to the two ventricles with silicone adhesive.[49] The patches were pneumatically driven to provide compression to the ventricles. This design allowed the two patch chambers to be controlled independently so that compression intensity and frequency could be customized for each ventricle.[28,50] An ECG was used to synchronize the movement of the device with the heart. Potential limitations of the HeartPatch include delamination of the patches from the cardiac surface and the rupture of the heart muscle.[41]
Soft artificial muscles (SAMs) have emerged as a potential actuation mechanism for many robotics and medical applications such as wearable assistive devices, surgical robots, and smart garments.[51–54] The motion and force of SAMs, depending on their materials and design architecture, can be controlled electrically, magnetically, thermally, chemically, and by pressure.[55,56] Their implementation in cardiac assist devices has been explored recently with impressive achievements.[9,41] The softness, conformability, and controllability of SAMs are desirable to make gentle contact with the heart surface as well as providing active compression. Roche et al. developed a soft robotic sleeve that wraps around the epicardium to assist compression.[57,58] The device has a cup shape made of two layers of pneumatic artificial muscles (PAMs) coated with silicone. One layer has the PAMs arranged in the circumferential direction to produce radial compression while the PAMs of the other layer are aligned axially with a twist to induce twist motion. The two layers of the soft robotic sleeve can be controlled independently, enabling the implementation of complex control strategies to mimic the heart's natural movements. The device performed better when operated with the combination of compress and twist than compress or twist alone. However, this soft robotic sleeve is relatively large with a thick wall, which is not favorable for implantation. In addition, it requires a complex control system and a bulky air source to operate the device. Kongahage et al. introduced a new extracorporeal VAD made from electrothermal actuators coated with silicone that could produce pulsatility instead of the continuous flow generated by conventional VADs.[59] The author also presented the design of a DCCD using the proposed soft thermal actuators to enclose and assist heart compression. However, this technology is limited to the use of heating- and cooling-driven muscles, which are slow and unsafe when working in the living body.
CorInnova DCCD (CorInnova, USA) is a double-walled cup-shaped device for the treatment of acute HF.[60] A series of nitinol wires separate the two silicone walls (or layers). The cavity between the two layers is filled with compressed air to create compression during systole. The space between the inner layer and the heart is filled with liquid to improve the interaction between the device and the heart.[61] The device is capable of multifunctional assistance including active compression and passive constraint.[62] Interestingly, the CorInnova device can be implanted with minimally invasive surgery, causing less postoperative pain and shorter hospital stays for patients. However, the CorInnova device has limited motion (squeeze by inflating the air chamber) so it cannot replicate the natural movements of the heart. Therefore, the CorInnova device may not be suitable for long-term implantation. More details on the current state and future direction of artificial muscles and soft robotics cardiac assist devices for the treatment of HF can be found in other studies.[63,64]
New Robotic Cardiac Compression DeviceIn this study, we introduce a robotic cardiac compression device (RCCD), which is a DCCD for the treatment of advanced HF. The primary drive of this study is based on the clinical need for a simple and effective cardiac assist device, which is expected to address some of the aforementioned limitations of current DCCDs. Its main actuation core is constructed from our previously developed miniaturized fluid-driven artificial muscle filament (AMF), which is flexible, responsive, and has a high length-to-diameter ratio and high elongation.[65] A special arrangement of AMFs, similar to the cardiac muscle fibres, allows the RCCD to simultaneously produce spatial motion including radial, axial, and torsional, mimicking the natural motion of the heart. The device requires the use of a small fluid volume that passes through a small driveline to operate, potentially enabling minimally invasive implantation. We measure several aspects of RCCD performance such as stroke volume, flow rate, compression force, and output pressure. An empirical model is proposed to describe the 3D motion of the RCCD. We also introduce an artificial pericardium (APC) and integrate it inside the RCCD. This new artificial component contributes to relieving stress on the heart surface with uniform force distribution and provides real-time pressure sensing for monitoring the heart–device interaction. Furthermore, we also introduce a modified RCCD that has an extra AMF to enhance the original RCCD performance. Finally, we experimentally demonstrate the RCCD performance on a fresh porcine heart.
Design and FabricationThis section introduces the design concept, working principle, and fabrication process of the RCCD, a type of DCCD that surrounds the failing heart to actively assist its contraction and pump sufficient blood to the body. We aim to create an RCCD that replicates 3D motion (radial, axial, and twisting motion in 3D space) of the real heart using a special biomimetic arrangement of our developed soft hydraulic artificial muscles. The RCCD is a semispheroidal shell consisting of an inner silicone elastomer layer and an outer actuation layer of AMFs (Figure 1a). To enable soft contact with the heart surface for safe operation, an inner soft layer is used. Two different designs for the inner layers will be employed in this work. The first prototype is a thin, low-Young's-modulus silicone elastomer that can stabilize the outer layer structure. The second prototype is a cavity-based soft silicone structure that is filled with saline water. This soft structure serves as an APC to provide uniform compression force distribution to the heart surface and real-time measurement of compression pressure via its fluid chamber (Figure 1b).
Enlarge this image.Figure 1. RCCD. a) Design concept. RCCD is a semispheroidal shell made of two layers of specially arranged AMFs. b) Design structure of the soft interface. (i) Conventional solid silicone interface. (ii) Pericardium-inspired soft interface with fluid cavity.
The outer layer is formed by two AMFs, including a compress-AMF inside and adjacent to a twist-AMF. The compress-AMF is helically wrapped around the silicon layer to induce radial motion. The twist-AMF is folded into multiple segments arranged along the RCCD axis of symmetry with a predefined twist to provide axial and twist movements. The twist-AMF segments resemble a left-handed helix, similar to epicardial fibers. A special feature of our design compared to existing state of the art is the use of a single and long soft hydraulic artificial muscle to form a biomimetic sleeve while offering small and constant hydraulic volume (<3 mL) during the operation and a miniature driveline (outer diameter, OD ≈2 mm) which connects the SAM layer to an external driving hydraulic source. The use of a small driveline potentially avoids high blood loss and infection risks in minimally invasive implantation while small fluid volume offers driving source minimization.
The compress-AMF and twist-AMF are thin, long, and flexible fluid-driven artificial muscles made by inserting a silicone tube inside an extension coil spring as a constraint layer. One end of the AMF is blocked while the other end is connected to a hydraulic source via a flexible fluid transmission tube. This simple fabrication method of insertion using commercially available components allows the AMF to have a high length-to-diameter ratio, high reliability, and repeatability, enabling low-cost mass production. When receiving positive hydraulic pressure, the AMF extends longitudinally due to the radial constraint of the coil spring on the silicone tube. Conversely, when the pressure is released, the AMF contracts and returns to its original length. The AMF is responsive, high speed (up to 20 Hz as reported in our previous study), and durable as well as has high contraction force and high energy efficiency.[65–70] In particular, it has a high elongation, meaning that the RCCD can be expanded to adapt to different heart sizes.
Using a special configuration of AMFs within the RCCD structure, it allows the transformation of the AMFs’ longitudinal motion into the RCCD spatial motion that mimics the natural heart motion. Specifically, when pressure is applied, the compress-AMF's elongation is transformed into the RCCD's radial expansion while the twist-AMF's elongation is converted into the RCCD's axial expansion and twisting motion in one direction. This pressurizing phase of the RCCD corresponds to the diastolic phase of the heart when the heart muscle relaxes to allow blood to passively fill the atria and partially fill the ventricles. Subsequently, when pressure is withdrawn, the contraction motion of the two AMFs generates the RCCD's radial and axial contraction and twisting motion in the reserve direction. This releasing phase of RCCD coincides with systole when the heart contracts to pump blood out of the ventricles (Figure 2a). Thereby, two working phases of RCCD complete the cardiac cycle.
Enlarge this image.Figure 2. Working principle of the RCCD. a) RCCD expands during diastole and compresses during systole to assist the failing heart to pump blood out of the ventricles. b) RCCD prototype made from AMFs and silicone elastomer.
An RCCD prototype is fabricated using the following process. First, we created a thin silicone layer from Ecoflex 00-30 (Smooth-On, Inc, USA) using casting molding technique. A set of molds made of polylactic acid (PLA) were manufactured using a 3D printer (Ultimaker, Netherlands). The silicone layer prototype had a semispheroidal shape with a base diameter of 76 mm, a height of 78 mm, and a thickness of 1.5 mm. Second, we made two AMFs with the same specifications (e.g., outer diameter of 2.5 mm) except for their lengths, where the compress-AMF had a length of 1000 mm and the twist-AMF was 940 mm long. Both AMFs were connected to the same flexible fluid transmission tube (or driveline, outer diameter 2 mm × inner diameter 1.2 mm × length 1200 mm, thermoplastic polyurethane) via a T-connector so they could be pressurized at the same time. Detailed specifications of the prototypes can be found in Table 1. Third, we pressurized (using degassed water) the two AMFs to 35% strain (input pressure around 1.2 MPa), then helically wrapped the compress-AMF around the silicone layer (with its inner mold attached), and arranged the twist-AMF on top of the compress-AMF. A higher elongation can be used to achieve a larger size of the sleeve with a stronger contraction force. The intersection points of the two AMFs were secured and attached to the silicone layer by elastic strings. It is worth noting that the silicone layer was molded with designated grooves on its exterior to facilitate the arrangement of AMFs. Finally, we removed the inner mold to obtain the RCCD prototype (Figure 2b). The result was a base diameter of 68 mm, a height of 72 mm, and a thickness of 7 mm at the releasing phase.
Table 1 Specifications of prototypes
a)OD: outer diameter, ID: inner diameter, L: length, k: spring constant, E: Young's modulus
Experimental Section Spatial Motion of RCCDThe RCCD was designed in a biomimetic manner so that it can induce spatial motion including radial expansion εr, axial expansion εz, and twisting motion with an angle change δω when being pressurized and vice versa (Figure 3a). To capture RCCD motion and to measure the shape change corresponding with input pressure, an experimental platform was built (Figure 3b). The platform consisted of a fixture to which the base of the RCCD prototype was attached so that its apex (or tip) could move freely. Two side cameras (Nikon, Japan and Nano Shield, Australia) 90° apart were used to capture RCCD dimension change in the radial and axial directions and a bottom camera (Nano Shield, Australia) was used to capture the twist angle change. Although one side camera can track the surface deformation, two cameras helped to minimize errors caused by the asymmetrical shape of the RCCD prototype during fabrication. A motorized linear slider (Zaber, Canada) was used to generate movements of a 3 mL syringe (BD Biosciences, Canada) to pump fluid (degassed water) to the RCCD.
Enlarge this image.Figure 3. Characteristics of the RCCD. a) Spatial motion illustration. b) Experimental setup using three cameras for motion capture. c) Side view shows axial and radial expansion at different input pressures. d) Bottom view shows the change of twist angle. e–g) Experimental results show the relationship between input pressure and axial expansion, radial expansion, and twist angle change.
We applied a step signal comprising ten steps forward with 4.4 mm increments and ten steps backward with 4.4 mm decrements where each step was 3 s apart from each other to the motorised linear slider. The slider speed was set to 10 mm s−1. The shape change of the RCCD at each step was captured by the three cameras and subsequently underwent an image processing process to calculate dimension changes. A pressure sensor (Honeywell, USA) connected to the fluid transmission tube was used to measure the input hydraulic pressure. The experiment was run five times. Experimental data were presented as mean ± standard deviation.
Figure 3c–d shows the side view and bottom view of the RCCD at different input pressures. As the input pressure increased, the RCCD expanded in radial, axial, and twisted clockwise directions. In contrast, when the input hydraulic pressure decreased, the RCCD radially and axially contracted, and twisted counterclockwise. The RCCD had a similar twisting motion as the native heart. Specifically, during systole (ejection), as viewed from the apex, the base rotated clockwise and the apex rotated counterclockwise.[71] The magnitudes of the radial, axial, and twisting movements varied slightly at different latitudes with an increase from base to apex.
The base of the RCCD experienced the least dimension change because it was tightened to the fixture. Figure 3e–g shows the experimental results which are the relationships between three output parameters (axial expansion, radial expansion, and twist angle change) and the RCCD input pressure at the latitude of the compress-AMF starting point (near the apex, marked with a cyan dot in Figure 3c). For example, under an input pressure of 1.13 MPa (8476 mmHg), the RCCD could induce an average radial expansion of 34.5%, an axial expansion of 14.5%, and a twisting angle change of 32.2°. All three charts show a clear nonlinear hysteresis relationship between the output parameters and input pressure, where a gap exists between the pressurizing and releasing profiles. Furthermore, the charts revealed the fast responsiveness of the RCCD evidenced by the absence of backlash, meaning that any change in input pressure causes a corresponding change in the RCCD motion.
Performance of RCCDWe set up another testing platform to measure the performance of the RCCD in terms of stroke volume, flow rate, and output pressure (Figure 4a). The RCCD prototype enclosed a silicone rubber univentricle and was attached to a fixture. A flexible tube connected the univentricle to a reservoir through an ultrasonic flowmeter (Atrato, Titan Enterprises, UK). This connecting tube served as the inlet and the outlet of the univentricle. The reservoir was raised ≈350 mm above the RCCD fixture to create initial hydraulic pressure. The ultrasonic flowmeter is capable of measuring instantaneous flow rate and cumulative volume of liquid flowing through its measuring channel. A pressure sensor was located on the connecting tube just 50 mm above the univentricle to measure output pressure. Similar to the previous experiment, the RCCD was actuated by a motorized linear slider via a 3 mL syringe with a pressure sensor attached to the input fluid transmission tube.
Enlarge this image.Figure 4. Performance of the RCCD. a) Experimental setup to measure flow rate and output pressure. b) Stroke volume and average flow rate of RCCD at different speeds. c) Flow rate at 0.2 Hz. d) Output pressure at 0.2 Hz. e) Compression force measurement setup. f) Force–pressure hysteresis at different loading pressures. g) Maximum force at different loading pressures. h) Relationship between blocked output pressure and input pressure. i) Maximum blocked output pressure at different loading pressures. It is noted that bpm is beats per minute.
To induce the RCCD motion, we applied different frequencies to the motorized linear slider using sinusoidal signals with an amplitude of 22 mm. For example, the RCCD could produce an average stroke volume of 107.6 mL at 0.1 Hz (or a rate of 6 beats per minute (bpm)), which is equal to an average flow rate of 0.65 L min−1 (Figure 4b). At the higher speeds of 12 bpm (0.2 Hz) and 15 bpm (0.25 Hz), the RCCD achieved average stroke volumes of 82 and 70 mL, which were equivalent to average flow rates of 0.98 and 1.05 L min−1, respectively. It is worth noting that a higher amplitude and frequency could enable a larger stroke volume and flow rate, respectively. We previously demonstrated that the soft hydraulic artificial muscle could achieve at least 240% elongation and a frequency of at least 20 Hz,[65] which could significantly increase the stroke volume and flow rate of the RCCD. However, our current motorized linear slider had a limitation in the stroke and frequency threshold, which was not able to simultaneously exceed 25 mm in amplitude and 3 Hz in frequency at the free load condition. In the case of the RCCD, due to the high force requirement for the motorized linear slider to drive the syringe plunger, the frequency limit was significantly reduced (0.25 Hz). In future work (out of this article's scope), we planned to use multiple DC motors to control multiple smaller syringes (e.g., 0.5 mL) connected to the soft sleeve, which overcame the difficulties in achieving high frequency.
Figure 4c shows the relationship between instantaneous flow rate and input pressure at 0.2 Hz. When the input pressure increased, the RCCD expanded, allowing water from the reservoir to be filled in the univentricle with a peak instantaneous flow rate of around 2.5 L min−1. Subsequently, when the input pressure decreased, the RCCD contracted to pump water from the univentricle to the reservoir with a maximum instantaneous flow rate of 2.8 L min−1. The output pressure (univentricular pressure) was inversely proportional to the input pressure (Figure 4d). The output pressure reached a minimum of 16.3 mmHg when the RCCD fully expanded at an input pressure of 1.2 MPa (9001 mmHg), whereas when the RCCD completely compressed, it produced a maximum output pressure of 50 mmHg. A hysteresis gap exists between the pressurizing and releasing phases of the input–output pressure chart.
We also characterized the compression force and blocked output pressure generated by the RCCD. We used a two-piece hoop made by 3D printing with a load cell (Futek, USA) sandwiched between them to collect compression force (Figure 4e). The hoop shape resembled the RCCD shape (semi-spheroid) for better contact. We pressurized the RCCD to a certain expansion before wearing it on the hoop and then reduced the input pressure using a 0.1 Hz sinusoidal signal with an amplitude that kept the minimum input pressure at 0.1 MPa. Three different loading pressures were used to accommodate three different sizes of hoop. The compression force has an inversely proportional relationship with the input pressure, in which the compression force is almost zero at the loading pressure and peaks when the input pressure drops to the lowest level (0.1 MPa) (Figure 4f). Figure 4f also reveals a modest hysteresis between the pressurizing and releasing phases of the pressure–force relationship. Three hysteresis profiles, corresponding to three loading pressures, share similar hysteresis patterns. With the loading pressure of 0.6, 0.8, and 1 MPa, when the input pressure was withdrawn, the RCCD achieved a maximum compression force of 10.5, 18.7, and 25.3 N, respectively (Figure 4g). These compression force values served as a quantitative reference for future cardiac compression devices.
The experimental setup to measure blocked output pressure was similar to that shown in Figure 4a, but the univentricular outlet was blocked instead of connecting to the reservoir. We pressurized the RCCD with a certain loading pressure and had the univentricle fully filled with water and then blocked the outlet. Similar to the compression force test, we reduced the input pressure using a 0.1 Hz sinusoidal signal with an amplitude that kept the minimum input pressure at 0.1 MPa. The blocked output pressure was inversely proportional to the input pressure with a very small hysteresis gap (Figure 4h). Corresponding to the loading pressure of 0.6, 0.8, and 1 MPa (4500, 6000, and 7501 mmHg), the RCCD generated blocked output pressure of 59.7, 62.6, and 69.2 mmHg, respectively, when the pressure was released (Figure 4i).
Mathematical ModelWe introduced here a mathematical model to describe the spatial motion of the RCCD, including radial expansion, axial expansion, and twisting angle change corresponding to the input pressure. Due to the complexity of modeling hyperelastic components, we proposed an empirical model based on the experimental calibration of AMF samples rather than theoretically building it from scratch. First, we established the relationship between output elongation and input pressure of three AMF samples (same outer diameter but different lengths) (Figure 5a). A longer AMF required a larger input volume to reach a certain threshold of elongation (Figure 5b). Interestingly, the pressure–elongation hysteresis profiles of the three AMFs almost overlapped with each other with a maximum standard deviation of 0.53% (Figure 5c). The AMFs achieved an elongation of 36.8% ± 0.25% when an input pressure of 1.2 MPa was applied. Next, we studied the geometric arrangement of the AMFs to establish the distribution of their longitudinal elongation to each constituent of the RCCD spatial motion (Figure 5d).
Enlarge this image.Figure 5. Mathematical model of the RCCD. a) Three AMF prototypes of different lengths are created for calibration. b–c) Elongation–volume and elongation–pressure relationships of the three prototypes. d) Geometric illustration of RCCD with spatial displacement. e–g) Comparison between mathematical models and experimental results on radial expansion, axial expansion, and twist angle change. AMF, artificial muscle filament; RMSE, root mean square error; P2PE, peak-to-peak error.
Finally, the elongation distribution and calibration data were used to infer the relationship between the RCCD spatial motion and the input pressure. Figure 5e–g shows the results of the mathematical model and the experimental data of the RCCD spatial motion. Despite the simplification of the model, it showed very good performance when following the experimental data with a small RMSE of 0.95% for radial expansion, 0.37% for axial expansion, and 1.6° for twist angle change. This empirical model was built upon calibration so it performed well in capturing the trend of experimental data. However, it has the limitation of loosely following local changes, as evidenced by a relatively large peak-to-peak error (P2PE) of 3.81% for radial expansion, 1.02% for axial expansion, and 3.8° for twist angle change. A detailed derivation of the mathematical model can be found in the Supporting Information.
Ex Vivo Results Uniform Force Distribution and in Situ Pressure Sensing with Artificial PericardiumThe RCCD is not only able to induce compression motion to the heart but also has the versatility to incorporate other features into its structure for additional functionality. Inspired by the pericardium that surrounds the heart to protect and lubricate the heart's movements,[72] this section introduces a double-walled sac known as APC sandwiched in between the RCCD and the heart (see Figure 1b and 6a). The APC consists of two thin layers made from a combination of silicone elastomer and nonstretchable fabric. This combination makes the layers soft, flexible, and nonstretchable, allowing them to safely and uniformly convey compression force from the RCCD to the heart compared to the normal silicone layer approach. The cavity between the two layers is called the pericardial cavity which is filled with liquid. The APC contributes two more functions to the RCCD. First, it provides gentle contact with the epicardial surface of the heart and helps distribute compression evenly to eliminate local stress that may cause damage to the myocardium. Secondly, the APC pericardial cavity pressure can be utilized as a sensing feature to monitor compression force.
Enlarge this image.Figure 6. Artificial pericardium (APC). a) APC structure. b) Fabrication process. (Left panel) The APC is made of a silicone–fabric combination, filled with water, and sandwiched between the RCCD and the heart. (Right panel) Comparison between the APC and normal silicone layer as soft interface. The APC offers full contact between the heart and the device for uniform force distribution while the normal silicone layer has nonuniform contact with gaps. c–e) Characteristics of APC show a relatively linear relationship between compression force and cavity pressure. f) Ultrasound setup to examine contact between RCCD and porcine heart. g) Ultrasound image of RCCD with APC showing continuous contact between APC and the heart. h) Ultrasound image of RCCD without APC showing local contact of RCCD and the heart. RMSE, root mean square error; AMF, artificial muscle filament.
Figure 6b shows the manufacturing process to create an APC prototype and its performance compared to the normal silicone layer approach. The inner and outer APC layers are semispheroidal shells made of Ecoflex 00-30 (Smooth-On, USA) and satin fabric using the molding technique. In detail, we made a semispheroidal shell from a piece of satin fabric and put it on a male mold and then dipped it in a female mold filled with uncured Ecoflex 00-30.
The prototype specifications can be found in Table 1. Next, we assembled the two APC layers and sealed their bases together with Sil-Poxy (Smooth-On, USA) to create a sac. A fluid transmission tube was connected to the cavity of the fully assembled APC at its apex to supply fluid and measure cavity pressure (note that a pressure sensor was connected to the fluid transmission tube). Finally, the APC was placed inside the RCCD so that the APC outer layer was in contact with the RCCD inner silicone layer. The integration was reinforced with sewing thread and Sil-Poxy.
We also investigated the sensing ability of the APC using the hoop stress setup shown in Figure 4e. The RCCD with integrated APC was pressurized and worn on the hoop and then the input pressure was withdrawn using a 0.1 Hz sinusoidal signal with an amplitude that kept a minimum pressure at 0.1 MPa. The relationships of input pressure, cavity pressure and force are presented in Figure 6c–e. The cavity pressure is inversely proportional to the input pressure with a fairly linear hysteresis profile. A hysteresis gap separates the pressurizing and releasing curves (Figure 6c). Similarly, the compression force has an inversely proportional relationship with the input pressure with a decent linear hysteresis profile. Interestingly, the hysteresis chart shows a very narrow gap between the two working phases (Figure 6d). The RCCD was pressurized to 0.64 MPa (4800 mmHg) and then reduced to 0.1 MPa to produce a cavity pressure of 45 mmHg and a compression force of 13.1 N. Figure 6e shows the proportional relationship between the compression force and the cavity pressure. Although a hysteresis gap exists, the hysteresis profile is fairly linear as evidenced by a relatively small RMSE of 0.5 N between the linear fitting curve (F(P) = 0.3646 P − 3.2713) and the experimental data. The experimental results demonstrated the feasibility of using the APC as a sensing element to monitor the compression force of the RCCD on the heart.
We also built an experimental setup to examine the APC's ability to provide soft contact and local stress relief to the epicardium. The setup consisted of a diagnostic ultrasound system (Mindray TE7, Shenzhen Mindray Bio-Medical Electronics, China) with an ultrasound probe (transducer) pointed toward the RCCD surrounding a fresh porcine heart (Figure 6f). The porcine heart was coupled with a rotary fixture along the heart axis. The entire setup was submerged in a bucket of water. We examined the contact between the heart's exterior and the RCCD with and without APC. The resulting ultrasound images are shown in Figure 6g,h. The experimental results revealed that the APC provided continuous contact with the heart. Furthermore, the liquid-filled APC cavity acted as a damping layer to prevent the local stress generated by the compress-AMF of the RCCD (Figure 6g). In contrast, local stress points were formed at the place where the compress-AMF made contact with the heart in the case of the RCCD without APC (Figure 6h). The silicone layer of the RCCD may provide a little stress relief but it is too thin and soft to distribute the compression force evenly across the heart's surface.
Customization for Performance EnhancementThe RCCD construction is highly customizable to provide the desired motion and force on the heart. For example, the two AMFs can be made longer to create a denser RCCD structure to increase compression. The predefined twist angle of the twist-AMF segments can be adjusted to manipulate the distribution of axial motion and twist angle change. This section investigates the enhancement of RCCD performance in terms of flow rate, stroke volume, compression force, and output pressure when incorporating an additional AMF in its structure. The additional AMF is 755 mm long and has the same specifications as the two original AMFs (Table 1). It is helically arranged 4.5 revolutions around the silicone layer, beneath the twist-AMF similar to the original compress-AMF (Figure 7a). The intersection points between the additional AMF and the twist-AMF are tightened with elastic strings. The inlets of the three AMFs are interconnected to receive a single source of hydraulic pressure. The modified RCCD is named configuration B in order to distinguish it from the original RCCD, which is configuration A. The design concept and prototypes of the two RCCD configurations are shown in Figure 7a. Configuration B has a denser structure than configuration A.
Enlarge this image.Figure 7. Customization of the RCCD for performance enhancement. a) Configuration A is the original RCCD and configuration B is the configuration A with an additional AMF. b–g) Performance comparison between two RCCD configurations in terms of flow rate, output pressure, stroke volume, average flow rate, maximum compression force, and maximum blocked output pressure.
The RCCD configuration B has undergone a similar series of experiments that we have performed with configuration A. The results showing the performance of the two configurations are aggregated and presented in Figure 7b–g and Table S3, Supporting Information. With sinusoidal signals that produce a maximum input pressure of 1.2 MPa, the RCCD configuration B generates higher instantaneous flow rates (25% enhancement) and output pressure (17.8% enhancement) than configuration A. The enhancement in stroke volume and average flow rate depends on the actuation speed with the highest increase of 23.1% at the lowest speed of 6 bpm. At the rates of 12 and 15 bpm, the increases are 20.5% and 16.1%, respectively. The compression force has the largest enhancement, reaching 73.3% at a loading pressure of 0.8 MPa. This loading pressure also enhances the blocked output pressure by 22%. With a loading pressure of 0.6 and 1 MPa, the compression force increases by 44.8% and 63.6% while the blocked output pressure increases by 19.1% and 20.5%, respectively. The experimental results clearly demonstrate the performance enhancement of the RCCD configuration B compared to configuration A.
Performance Demonstration on Porcine HeartIdeally, the RCCD should be tested on the hearts of live animals that have been chronically induced with HF because it is designed to support a failing (still working) and hypertrophic heart instead of a dead one. However, ex vivo experiments on a dead heart still provide important insight into the heart–device interactions and the overall functions and performance of the device. In this article, we used a porcine heart obtained from a local supermarket (Woolworths Supermarkets, Randwick, Sydney, Australia). Such preliminary results are crucial for device improvement and preclinical trials in the future. We implemented several variants of the RCCD on a deceased porcine heart to examine their performance in generating motion, flow rate, and output pressure.
We supplied positive hydraulic pressure to expand the original RCCD (configuration A) and wore it on a porcine heart (Figure 8a). The RCCD apex was in contact with the heart apex and the RCCD enclosed the entire epicardial surface of the heart from apex to base. The RCCD silicone layer was directly contacted with the epicardium. We adjusted the input pressure of the RCCD to achieve a firm wrap on the heart. The left and right ventricular chambers of the porcine heart were equipped with elastic balloons connected to a fluid tube, followed by a flowmeter and a water reservoir. A pressure sensor was attached to the fluid tube to measure output pressure. We then withdrew the input pressure using a 0.2 Hz (12 bpm) sinusoidal signal with an amplitude that kept a minimum input pressure at 0.1 MPa. The maximum input pressure was 1.2 MPa.
Enlarge this image.Figure 8. Performance demonstration of the RCCD on a porcine heart. a) RCCD configuration A surrounds the heart to provide compression that produces b) flow rate and c) output pressure. d) The heart is wrapped with RCCD configuration A with an integrated APC. e) RCCD configuration B compresses the heart to produce stroke volume. f) The heart is wrapped with RCCD configuration B.
The relationship between instantaneous flow rate and input pressure is shown in Figure 8b, which is similar to the profile when the RCCD pumped an artificial univentricle (Figure 4c). However, the flow rate magnitude is smaller in the case of pumping the porcine heart, which is 2 L min−1 pump out and 1.5 L min−1 fill in (peak instantaneous flow rates). The difference between pump-out and fill-in flow rates is due to two distinct working phases of the device. In the compression phase, the device actively compresses the heart to pump blood out of the ventricles, resulting in a higher instantaneous flow rate. In contrast, the device expands to let the blood fill the ventricles passively in the expansion phase, leading to a lower instantaneous flow rate. At the rate of 12 bpm, the RCCD coupled with the porcine heart produced a stroke volume of 61 mL, equivalent to an average flow rate of 0.73 L min−1. It also generated an output pressure of 39.5 mmHg when the input pressure was completely withdrawn (Figure 8c). A large hysteresis gap between the pump-out and fill-in phases can be seen in the input–output pressure hysteresis chart. We observed that the RCCD provided the porcine heart with spatial motion including radial, axial, and torsional simultaneously during both compression and expansion.
We also implemented the original RCCD with integrated APC on the porcine heart (Figure 8d). One obvious improvement is that the APC acted as a protective layer so the AMFs did not compress directly into the epicardium. As a result, the RCCD with APC experienced a smaller slip (displacement < 2 mm) on the cardiac surface compared with the RCCD without APC (14 mm displacement) after performing 15 cycles of compression and dilation. In addition, the APC cavity pressure offered another means of monitoring compression besides the input pressure. When the input pressure is completely withdrawn, the APC cavity pressure reached 58.5 mmHg. Although the RCCD exerted spatial motion, we were unable to observe the heart movements due to the opaque APC. Furthermore, the RCCD with APC underperformed RCCD without APC, evidenced by an average flow rate reduction of about 18% and a decrease in output pressure by 23%. The underlying mechanism of this reduction comes from the reaction force of APC fluid to the RCCD. Specifically, besides transferring compression force from the RCCD to the heart, APC fluid also stretched the APC outer wall at the mesh eyes of the RCCD, reducing the efficiency of compression force transmission and thus decreasing flow rate and output pressure. The presence of air bubbles inside the APC cavity due to imperfect APC fabrication also contributed to the reduction in compression efficiency.
The last RCCD variant that we implemented on the porcine heart was the modified RCCD (configuration B). Similar to configuration A, the RCCD configuration B was pressurized and worn on the porcine heart to induce motion. Figure 8e shows two working phases of the RCCD configuration B coupled with the heart, where stroke volume is indicated by the level difference of the red-dyed water column. A closer image of the contact between the device and the heart is shown in Figure 8f. There is a thin silicone layer between the AMFs and the epicardium so the motion generated by the RCCD is well transferred to the heart. With an actuation speed of 12 bpm, the RCCD configuration B helped the heart produce an average flow rate of 0.83 L min−1 and an output pressure of 43.6 mmHg, which means an increase of 13.7% and 10.4%, respectively, compared to configuration A. These results consolidate the previous conclusion that configuration B outperforms configuration A.
DiscussionDriven by the large clinical need and limitations of current cardiac assist devices, this study introduced a new DCCD called RCCD to assist in the treatment of HF. The proposed RCCD is designed to surround the epicardium to actively provide compression to pump blood out of the ventricles during systole. It then expands during diastole, allowing blood to fill the ventricles. Active diastolic support is feasible assuming adhesion between the RCCD and cardiac surface has been established. Inspired by the structure of cardiac muscles, especially the myocardial sheet and fiber arrangement from endocardium to epicardium, the RCCD has a semispheroidal shape made from a special arrangement of two AMFs, including a spirally arranged AMF to induce radial motion and an axially arranged AMF to induce axial and twisting movements. This design allows the RCCD to produce spatial motion including radial, axial, and torsional movements, mimicking the natural motion of the heart. Previous researchers have highlighted that a cardiac assist device that conforms to the natural mechanics of the heart motion would be preferable for the treatment of HF.[58] Additionally, when mimicking the natural movement of the heart, DCCDs will well support the heart without impeding its natural movement, potentially helping already weak heart muscles.
The RCCD expanded by 34.5% radially, 14.5% axially, and twisted by 32.2° when receiving positive hydraulic pressure. Subsequently, it contracted and twisted in the opposite direction when the input pressure was withdrawn. The twist angle achieved by the RCCD is larger than the maximum systolic torsion angle of the normal heart, around 8° ± 2.1°.[73] However, the resulting twist angle can be adjusted by varying the predetermined twist angle of the AMF segments during fabrication. Studies of total heart volume variation (THVV) have reported conflicting results ranging from 5% to 13%, including recent work using cine magnetic resonance imaging (cine-MRI) by Carlsson et al. showed THVV of 8% between diastole and systole in healthy subjects.[74] The study also showed a small change in the length of the heart (axial contraction) of 0.9% ± 0.5%. In another study, Carlsson et al. concluded that more than 80% of THVV during the cardiac cycle is due to radial changes.[75] Therefore, the radial and axial changes achieved by our device are much larger than those of the normal human heart. However, these overperformances should not be a problem since the dynamic interaction between the device and the heart helps to reduce these changes. We also expect the external compression force provided by the RCCD will compensate for the weak myocardium of the failing heart. The relationships between output movements and input pressure are nonlinear with relatively large gaps in their hysteresis profiles (Figure 3e–g). Nonlinear hysteresis is an inherent characteristic of AMFs constructed from silicone elastomers. The hysteresis gaps represent energy loss when the AMFs change their working phase from pressurizing (accumulating elastic energy) to releasing (discharging the stored energy). Interestingly, no backlash exists in the hysteresis profiles, which means that the RCCD movements are highly responsive to hydraulic pressure.
The RCCD could produce a stroke volume from 70 to 107.6 mL depending on the actuation speed, which is comparable to the soft robotic sleeve made by Roche et al. with 84 mL.[58] However, the soft robotic sleeve could perform at 80 bpm owing to multiple large PAMs while the RCCD achieved only 15 bpm. The use of long AMFs with a small inner channel of the silicone microtube (≈0.3 mm) has caused a limitation in the actuation speed, due to the microhydraulic friction and relaxation of the elastic materials. When the pressure increases, the AMF will elongate and when the pressure is decreased, the AMF will shorten. However, under high pressure, especially high frequency, the AMFs could not achieve the desired elongation threshold needed to induce a change in the sleeve volume, resulting in stroke volume reduction. From our experience, this problem can be solved if we reduce the length of the AMF and increase the internal diameter of the hollow silicone tube. It means that we can use multiple AMFs with a shorter length. In vitro experiments revealed that the RCCD could produce a maximum instantaneous flow rate of 2.8 L min−1, an output pressure of 50 mmHg, a compression force of 25.3 N, and a blocked output pressure of 69.2 mmHg. These results suggest that the RCCD underperformed compared to a normal heart but is comparable with a failing heart.[62] Therefore, the RCCD is intended to assist in the treatment of HF rather than taking over the heart function.
An empirical model based on experimental calibration has been developed to describe the 3D motion of the RCCD. Based on the input–output relationships established in the model, the resulting radial, axial, and torsional movements can be fine tuned by adjusting input parameters during the design and fabrication, such as the AMF lengths, the number of twisting segments, and the predetermined twist angle. The model performed well in capturing the tendency of radial expansion, axial expansion, and twist angle change with acceptable RMSEs. However, it was unable to closely follow every local change in the experimental data. Due to the experimental-based approach, the proposed empirical model can theoretically work with filament size variations (e.g., multiple shorter AMFs with larger diameters).
Inspired by the natural pericardium, we created a silicone APC located inside the RCCD to provide soft and conformable contact with the heart. The double-layer structure with a fluid-filled cavity helped distribute compression from the RCCD to the heart evenly without local stress, thus potentially preventing myocardial damage. The cavity pressure of the APC can be used to monitor compression force with a fairly linear relationship (Figure 6e). Although we were unable to provide quantitative data on the impact of the APC on the RCCD twist motion, it is more likely that the twist angle would be significantly reduced. However, that should not be a problem since the current RCCD could produce a much larger twist angle change (32°) compared with the maximum systolic torsion angle of the normal heart (≈8°). Besides, the APC design structure can be improved to make the torsional transmission more efficient such as introducing multiple semirigid links between the inner and outer layers of the APC. A modified RCCD with an extra helically arranged AMF was created to enhance the performance of the original RCCD. The modified variant showed significant improvement in all aspects including flow rate, stroke volume, output pressure, and compression force (Figure 7 and Table S3, Supporting Information). Besides providing additional compression, the extra AMF also made the RCCD structure denser (reducing gaps between active filaments), thus conveying movements more effectively.
For demonstration, we implemented several RCCD configurations on a dead porcine heart. Overall, all RCCD configurations were able to wrap around the porcine heart from base to apex and induce 3D movements. Coupled with the heart, the original RCCD produced a stroke volume of 61 mL, a cardiac output of 0.73 L min−1, and an output pressure of 39.5 mmHg. An enhancement from 10.4% to 13.7% was achieved by the modified RCCD. These results suggest that the performance of the RCCD is reduced when paired with the heart compared with an artificial univentricle. The underlying mechanism of these reductions includes the complex structure of the ventricular chambers that create air pockets when enclosing the elastic balloons. The compressible air pockets decreased compression efficacy. Another cause is that the RCCD slipped slightly on the heart surface during compression because the device was solely attached to the heart's exterior by friction rather than adhered by glue or sewn by sutures.
We used sinusoidal signals as a standardized input throughout the study to achieve consistency of device behavior for ease of comparison. The sinusoidal signal has the advantage of simple modulation and soft takeoff and landing characteristics, eliminating sudden changes in actuation speed and thus smoothing the output signals and improving the service life of prototypes. We are confident that the proposed device can withstand different types of input signals (e.g., triangular, trapezoidal) as long as they are within the speed limits of the motorized linear slider. The obtained hysteresis results (flow rate and output pressure versus input pressure) will probably look similar to those reported in this work with minor differences at the take-off and landing positions.
The proposed RCCD has left some limitations for future improvement. Multiple shorter AMFs with a larger diameter and stronger contraction can be used to construct a more advanced RCCD with faster actuation speed and stronger compression than the current version. Also, a faster and stronger linear slider is required to increase the device speed. A mathematical model is required to describe the enhancements needed to achieve faster actuation speed and stronger compression force. In addition, a denser structure of AMFs increases the transmission of motion from the device to the heart. Subsequent generations of the RCCD should have an effective binding method to maintain contact with the heart surface. Some potential adhesion approaches include the use of adhesive porous elastomers,[49] suction cups,[76] and biointegration of medical mesh.[42] These attachment methods can be performed through either open or laparoscopic surgery. Later stages of the development progress after establishing the device–heart adhesion have to include evaluation of force transmission efficiency. The current RCCD in this article is unable to provide the LVs and RVs with two different compression forces. One possible solution is to develop a cardiac compression device with different local stiffness or different rates of local contraction, leveraging the capability of our soft muscle that can tune its generated force and elongation at the time it is fabricated. Other types of input signals (e.g., triangular, trapezoidal) should be tested to investigate the device behavior. A control system with integrated ECG is required to synchronize the device's movements with cardiac motion.[42,58] We strongly recommend developing a compact and powerful wearable actuation unit to drive the RCCD. It is necessary to perform a durability test of several thousand cycles for future RCCD prototypes. Also, a safe failing mode of the device should be developed for patient safety. There is currently a lack of long-term clinical evidence to prove the safety and effectiveness of DCCD as well as the effect of DCCD on the papillary muscles and heart valve function. Therefore, preclinical trials in live animals and clinical trials in humans should be conducted to evaluate the safety, potential complications, and therapeutic efficacy of the device. From a clinical perspective, further studies are needed to evaluate the reduction in efficacy when the RCCD compresses both LV and RV simultaneously and the impact of RCCD compression on coronary circulation. Treatment recommendations such as long-term implantation or bridging therapy for heart transplantation should also be considered.
ConclusionThis study has introduced a proof-of-concept RCCD made of AMFs and silicone elastomers to assist in the treatment of advanced HF. The proposed cardiac assist device surrounds the epicardium to actively provide compression during systole and expansion during diastole. The device could generate spatial motion including radial, axial, and torsional movements to mimic the natural motion of the heart. An APC is developed to provide gentle contact with the heart and for sensing purposes. The RCCD structure can be customised to enhance performance. The proposed device is implemented on a dead porcine heart to induce 3D motion, output flow rate, and pressure. The design concept and preliminary results of this study are expected to inspire further improvement as well as extend the development of active cardiac assist devices.
AcknowledgementsThe authors acknowledge support from the UNSW Start-Up Grant (PS58173), the UNSW Scientia Fellowship Grant (PS46197), and the Vanguard Grant from the National Heart Foundation of Australia (RG204224).
Open access publishing facilitated by University of New South Wales, as part of the Wiley - University of New South Wales agreement via the Council of Australian University Librarians.
Conflict of InterestThe authors declare no conflict of interest.
Data Availability StatementThe data that support the findings of this study are available from the corresponding author upon reasonable request.
© 2024. This work is published under http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Details
1 Graduate School of Biomedical Engineering, Faculty of Engineering, University of New South Wales (UNSW), Sydney, NSW, Australia
2 Graduate School of Biomedical Engineering, Faculty of Engineering, University of New South Wales (UNSW), Sydney, NSW, Australia; College of Engineering and Computer Science, Vin University, Hanoi, Vietnam
3 Department of Cardiology, St Vincent's Hospital, Sydney, NSW, Australia; St Vincent's Clinical School, Faculty of Medicine, UNSW, Sydney, NSW, Australia
4 School of Mechanical and Manufacturing Engineering, Faculty of Engineering, UNSW, Sydney, NSW, Australia; Tyree Foundation Institute of Health Engineering (IHealthE), UNSW, Sydney, NSW, Australia
5 Graduate School of Biomedical Engineering, Faculty of Engineering, University of New South Wales (UNSW), Sydney, NSW, Australia; Tyree Foundation Institute of Health Engineering (IHealthE), UNSW, Sydney, NSW, Australia