- Email: [email protected]
Modelling the left ventricle using rapid prototyping techniques
Disponible en ligne sur
www.sciencedirect.com IRBM 34 (2013) 226–234
Original article
Modelling the left ventricle using rapid prototyping techniques B. Van Der Smissen a,b,∗ , T. Claessens a,b , P. Verdonck b , P. Van Ransbeeck c , P. Segers b a
University College Ghent, BioMech, 9000 Ghent, Belgium University College Ghent, IBiTech BioMMeda, 9000 Ghent, Belgium c University College Ghent, Department of Mechatronics, 9000 Ghent, Belgium b
Received 16 October 2012; received in revised form 28 March 2013; accepted 1st April 2013 Available online 13 May 2013
Abstract Biomechanical research of left ventricular function involves the assessment and understanding of both ventricular wall mechanics and deformation and intraventricular flow patterns, as well as how they interact. Experimental research using hydraulic bench models should therefore aim for an as realistic as possible simulation of both. In previous experimental investigations, wall deformation was studied by means of thin-walled passive experimental models, consisting of a silicone membrane in a closed box, which is squeezed passively by an externally connected piston pump. Although the pump function of these models has already been well established, the membrane deformation remains unpredictable and the effect of muscle contraction – and hence natural wall deformation – cannot be simulated. In this study, we propose a new design of an experimental hydraulic left ventricular model in which left ventricular wall deformation can be controlled. We built this model by a combination of rapid prototyping techniques and tested it to demonstrate its wall deformation and pump function. Our experiments show that circumferential and longitudinal contraction can be attained and that this model can generate fairly normal values of pressure and flow. © 2013 Elsevier Masson SAS. All rights reserved.
1. Introduction Heart failure is an increasingly important health problem in most developed countries [1]. An important focus in biomechanical research of congestive heart failure is the role of the deformation of the ventricular wall in the interaction with blood flow. Therefore, research using experimental models should aim to a realistic imitation of both. In previous experimental investigations, wall deformation was studied by means of thin-walled passive models, consisting of a silicon membrane in a closed box that is passively squeezed by an externally connected piston pump. Experimental models relying upon this deformation mechanism have been used to study the influence of the different determinants of left heart performance on transmitral flow [2], to test the hydrodynamic performance of heart valves [3–10] or to investigate left ventricular diastolic filling using colour M-mode Doppler echocardiography [11]. Although the pump function of these models is well established, the membrane deformation
remains unpredictable and the effect of muscle contraction – and hence natural wall deformation – cannot be simulated. 2. Objectives In this study, we propose a new design of an experimental hydraulic left ventricular model in which left ventricular wall deformation can be controlled. First, we have built this model by a combination of rapid prototyping techniques and after that tested it to demonstrate its wall deformation and pump function by measuring aortic flow and ventricular pressure. The model, which is described in this article, is the result of an optimization process in which we gradually improved our previous designs based on different design criteria (e.g. practical feasibility). However, in this article, we exclusively describe the design and composition of the final model that was also physically realized. 3. Physiological background
∗
Corresponding author. University College Ghent, BioMech, 9000 Ghent, Belgium. E-mail address: [email protected] (B. Van Der Smissen). 1959-0318/$ – see front matter © 2013 Elsevier Masson SAS. All rights reserved. http://dx.doi.org/10.1016/j.irbm.2013.04.001
The heart is a four-chambered, muscular organ that continuously pumps blood through the body’s extensive network of arteries and veins. The left ventricle (LV) can be considered
B. Van Der Smissen et al. / IRBM 34 (2013) 226–234
227
as its most important and powerful pumping chamber, and is most frequently studied [12]. The LV wall consists of three layers: the epicardium, the myocardium, and the endocardium. The endocardium is the thin inner layer of endothelium while the epicardium is an external visceral layer covering the heart. The myocardium, the middle layer, forms the bulk of the LV and is responsible for the actual contraction [13]. It consists of interweaving bundles of cardiac muscle fibres embedded in connective tissue structure and spirally arranged around the circumference of the heart [13,14]. As a result of this anatomical architecture and of the timing and sequence of electrical excitation, during contraction the LV deforms as follows: (a) the diameter of the LV decreases while (b) the base-apex distance decreases (c) in a rotating manner [14]. As such, global ventricular wall deformation can be basically described in terms of (a) circumferential contraction, (b) longitudinal contraction, and (c) a torsion motion [14].
4. Contraction principle Up till now, experimental left ventricular models have been primarily used to test the hydrodynamic performance of heart valves [3–10] and, to a lesser extent, to study transmitral flow [2] or left ventricular diastolic filling [11]. Given these specific foci of research, not much attention has been directed towards mimicking the complex wall deformation patterns. In all referred models, the ventricle has been modelled as a passive thin-walled silicone membrane within a closed chamber [3–11]. By using a piston pump, water or pressurized air is applied within the closed box and the internal membrane compresses which consequently results in LV volume reduction. Although it is well demonstrated that these passive models provide adequate pump function, the unpredictability of the membrane deformation does not allow simulating the effect of muscle contraction and hence natural wall deformation. In contrast to this thin-walled conception of the LV, we conceived the LV as a thick-walled hydraulic model composed of three layers. First is an inner thin-walled membrane, second is a middle layer that includes flexible bars, which we assume to control the wall deformation and third is an exterior shell. The contraction principle is as following. During the systolic phase, a piston pump pushes a certain amount of fluid into the region with flexible bars, compressing the internal membrane and stretching the flexible bars. In this way, the thin membrane is supported during the whole heart cycle and does not randomly deform as was the case with previous models. This results in wall deformation that is controlled by the flexible bars during systole while LV volume is decreasing. Likewise, during the diastolic phase, fluid is sucked from the region with flexible bars and the wall deforms back to its initial geometry, which also occurs in a controlled manner. These three layers of the ventricular part could be considered as analogues of the human heart. In such an analogy, the thin membrane represents the endocardium, the flexible bars connecting the membrane with a stiff supporting frame represent the myocardium that actually accounts for ventricular contraction,
Fig. 1. Illustrative drawing of a 16-segment left ventricle (LV) model representing the 16 segments, which is often used in echocardiography. The model consists of three slices: (1) the basal and (2) the mid-ventricular slices which are uniformly divided into six segments and (3) the apical slice which is split into four equal segments. Reproduced from Han, 2004 [15].
and the exterior shell covering the latter parts represents the epicardium, which is the external visceral layer of the heart. The degree of control over the deformation of the inner membrane is determined by the tension and number of flexible bars that are arranged over the membrane surface. The higher this number, the more accurate the control of the wall deformation will be. In our model, we provided the possibility to stretch or loosen the flexible bars by adding or removing modular discs in order to control the wall deformation more efficiently (Section 5.1). We have set the number of bars to 31, which represent the corners of a 16-segment model that is often used in clinical echocardiography to define LV pathologies [15]. The 16 segments are performed by dividing the LV model into three slices and each slice into several segments. As such, the basal and the mid-ventricular slices are uniformly divided into six segments and the apical slice into four segments. The configuration of the 16-segment model is illustrated in Fig. 1.
5. Design and composition of the left ventricle model This contraction principle was further optimized in terms of practical feasibility and resulted in the model shown in Fig. 2. In order to facilitate experiments, we modularly designed the model. It consists of two main parts: a ventricular part (Fig. 2b) and a valve housing (Fig. 2a). The ventricular part takes care of the ventricular contraction while the valve housing holds two (mechanical) heart valves and connects the model to an afterload configuration, e.g. a lumped hydraulic windkessel model of the vascular system.
228
B. Van Der Smissen et al. / IRBM 34 (2013) 226–234
Fig. 4. The eye pins are designed with (a) an eye to connect the flexible bars, (d) a perforated flange to assure embedding in a tight connection with the flexible thin membrane, (e) a centering pin to provide (c) a constant membrane thickness of 3 mm and (b) guiding planes to align them perpendicular to the membrane.
Fig. 2. The model consists of two main parts: (a) a valve housing and (b) a ventricular part.
5.1. Design features of the ventricular part
one end to the supporting points in the membrane, the so-called eye pins (Fig. 3g). At the other end, these bars are connected to the supporting frame by modular discs (Fig. 3i). We designed these eye pins (depicted in more detail in Fig. 4) with the following four features:
As depicted in more details in Fig. 3, the ventricular part consists of four components: the thin flexible membrane (Fig. 3a), the flexible bars (Fig. 3b), the stiff supporting frame (Fig. 3c) and the exterior shell (Fig. 3e). First, the thin membrane (Fig. 3a), the inner layer of the ventricular part, is modelled as a truncated ellipsoid with a volume of 130 mL, base-apex length of 85 mm, short axis diameter of 45 mm and with an overall wall thickness of 3 mm. The membrane is made of a 2-component flexible Polyurethane with an E-modulus of 5 MPa. We also provided the membrane with a sealing ring (Fig. 3f) in order to prevent leakage between the region with the flexible bars and the LV cavity. Second, the region with the flexible bars (Fig. 3b), connecting the inner and outer layer of the ventricular part, can be assumed as the middle layer. Importantly, these bars need at the one hand to be flexible enough to allow deformation of the membrane and, at the other hand, to be stiff enough to exert control over the wall deformation. Empirically, we found the following parameters appropriate to account for these requirements in our model, that is a rubber material with an E-modulus of approximately 5 MPa and a cross-section of 4 mm2 . The flexible bars are connected on
• an eye (Fig. 4a) to which a flexible bar finally is connected; • a perforated flange (Fig. 4d) to assure embedding in a tight connection with the flexible thin membrane; • a centering pin (Fig. 4e) to provide a constant membrane thickness; • guiding planes (Fig. 4b) to align them perpendicularly to the membrane (Section 6).
Fig. 3. The ventricular part consists of (a) a thin membrane including (f) a sealing ring, (b) flexible bars, (c) a supporting frame, (d) an O-ring, (e) an external shell, (g) eye pins, (h) supporting points, (i) modular discs and (j) connection to the piston pump.
Fig. 5. The modular discs are designed with (d) a hole and (a) a cut-out in which a transverse spill (Fig. 8c) fits to allow easy connection using different types of flexible bars, (c) a cut-out to facilitate tightening and loosening of the flexible bars and (b) an adjusting ring to fit them together (e).
The modular discs (Fig. 5) were designed also with several features: • a hole (Fig. 5d) where the flexible bars pass through; • a cut-out (Fig. 5a) in which a transverse spill (Fig. 8c) fits in order to allow easy connection by using different types of flexible bars; • a cut-out (Fig. 5c) to facilitate tightening and loosening of the flexible bars; • an adjusting ring (Fig. 5b) to keep the different modular discs together (Fig. 5e). The third component, the supporting frame (Fig. 3c), is located at the outer side of the middle layer of the ventricular part. It serves not only to keep the supporting points (Fig. 3h) in a fixed position, but it also enables visual inspection of the wall deformation as well as a more practically feasible assembly of the flexible bars. We designed its geometry according to
B. Van Der Smissen et al. / IRBM 34 (2013) 226–234
229
6. Rapid prototyping materials and methods
Fig. 6. The valve housing contains (e) a valve plane, (a) two gates and positions, (d) two mechanical valves. Leakage between valves and valve plane is prevented by (c) O-rings, which are pressed by (b) compression nuts.
several design criteria. First, the general geometry is set with a constant offset (25 mm) to the membrane and it consists of a coarse-meshed structure (beam height and width of 3 and 5 mm respectively). In addition, supporting points are included to realize a perpendicular orientation of the flexible sticks with respect to the membrane surface. As such, the flexible bars have a constant length and therefore ensure uniform control of the LV wall deformation. Finally, the exterior shell (Fig. 3e) functions as the exterior box in which fluid is pumped by a piston pump. Similar to the geometry of the supporting frame, the exterior shell is designed with a constant offset distance from the membrane. We have set an offset of 2 cm with respect to the supporting frame to allow stacking of several modular discs (Fig. 5c) intended for extrastretching of the flexible bars. Another feature of the exterior shell is the pump connection, which is located underneath at apex position (Fig. 3j). Leakage from the middle layer of the ventricular part (region with the flexible bars) towards the outside of the model is prevented using a sealing ring (Fig. 3d), which is located between the exterior shell and the valve housing. 5.2. Design features of the valve housing The valve housing assembly consists of a valve housing part (Fig. 6 and Fig. 9) that positions and seals two mechanical heart valves, one at mitral and the other one at aortic position (Fig. 6d and Fig. 9b and g). Each valve is oriented with an angle of 55◦ with respect to the long axis of the LV. Leakage between the valves and the valve plane is prevented using an O-ring (Fig. 6c and Fig. 9f) that is compressed by a compression nut (M42 with a pitch of 1.5 mm) (Fig. 6b and Fig. 9c and e). The valves are positioned as closely as possible to the valve plane in order to minimize the non-contractile space of the LV cavity. However, the limiting factor hereby is the minimal wall thickness of the middle part of the valve housing. This is the place where the distance between both valves is minimal (we set this distance to 2 mm). In order to position the valves even closer to the valve plane, we shaped the two canals with a constant surface area as following: the shape of the cross-section gradually changed starting from a circular-shape at the level of the valves, up to a semicircular-shape towards the LV cavity.
In the past two decades, rapid prototyping (RP) techniques have substantially changed and improved prototyping manufacturing practices in both industry and research. These techniques allow to quickly fabricate three-dimensional products using Computer Aided Design (CAD) systems [16], producing prototypes in a broad range of materials with outstanding precision, whilst taking only a couple of hours of production. These techniques are now successfully integrated into product development methodologies as they allow to address the market’s requirements in a customized way. Being continuously under development, the range of applications of RP technology now also expands to the biomedical field as it provides innovative support for research plans requiring prototypes [17]. RP technologies can be of assistance in every stage of the development process of novel devices, can help to address various problems arising during development and also supports design [17]. Current examples of rapid prototyping technology applications in the medical field include surgical planning, and the design and development of (patient-specific) prostheses [18], surgical tools, implants, and other biomedical devices [19]. In this study, the creation of the thick-walled LV model was made possible using an innovative combination of three different rapid prototyping techniques (Fig. 7), namely Selective Laser Sintering (SLS), Stereolithography (SLA) and vacuum casting. SLS and SLA, comprising two basic techniques of rapid prototyping (also popularly known as 3D printing) that are both additive manufacturing techniques. These techniques differ fundamentally from traditional manufacturing where an object is shaped by material removal, such as cutting, drilling, and milling. In the case of additive manufacturing, the method to produce shaped parts is by creating or adding solid material layer by layer [20]. The designed geometry of the desired object can be created using 3D drawing computer software or can be derived from scanning data of imaging technologies such as magnetic resonance imaging (MRI) or computed tomography (CT) techniques [21]. This geometrical construction is then virtually sliced into thin layers that correspond to the layer thickness provided by the layer by layer printing process [19]. SLA was introduced in the late 1980s and is one of the most powerful and versatile of all rapid prototyping techniques. It has the highest fabrication accuracy and a wide range of materials that can be processed [19]. The material used for SLA is a liquid photo-curable resin, with an ultraviolet laser emitting photons, which initiate the polymerization of monomers [16]. Based on this principle, solid objects are built in a tank of liquid resin by successively curing thin layers, one on top of the other. SLS is a powder-based layer-process. Laser beams are used as a heat source for scanning and fusing powder materials in predetermined sizes and shapes of layers. After the first layer is scanned, a second layer of loose powder is deposited over it, and the process is repeated from bottom to top until the object is complete [22]. One advantage of SLS is that it can be applied with a wide range of materials [23], such as wax, cermet, ceramics, nylon/glass composite, metal-polymer powders,
230
B. Van Der Smissen et al. / IRBM 34 (2013) 226–234
Fig. 7. This figure shows the practical realization of (a) our model, (b) the valve housing, (c) an assembly of the supporting frame and membrane without the flexible bars, (d) the external shell, (e) the supporting frame and (f) the membrane containing the embedded eye pins.
metals, alloys or steels [24,25], polymers [26], nylon, and carbonate [27]. As such, materials can be chosen based on their characteristics dependent on the type of process, and therefore, SLS is primarily used to manufacture functional parts. The third technique we used is vacuum casting, a manufacture process often used to produce small series of elastomer products. This process is not an additive manufacturing techniques (such as SLA and SLS), but a casting technique. It starts by placing a two piece silicone mould in a vacuum chamber. Then, the elastomer is mixed and degassed and subsequently poured into the mould. The vacuum is used to remove the entrapped air and to draw the liquid elastomer material into a silicone mould. After the vacuum is released, the mould is removed from the chamber. Finally, the liquid elastomer is cured in an oven and the mould is removed to release the completed casting. This silicone mould can be reused to produce identical copies of the needed elastomer parts. Application of RP techniques to build the LV model: the combination of these three techniques enabled us to build the complete model, which is composed of three different types of materials, such as: • epoxy (by SLA) to build the external shell; • polyamide (by SLS) to fabricate the supporting frame, valve plate and the tinier parts such as the eye pins;
• 2-component polyurethane (by Vacuum Casting) to produce the LV membrane. The external shell (Fig. 7d) was built from epoxy using the SLA technique because this technique allows semi-transparency in combination with solid properties providing enough strength to deal with the pressurized media. The SLS technique, however, was used to build parts that can resist relative high forces such as the modular discs (Fig. 8c), the supporting frame (Fig. 7e), the valve plane (Fig. 9) and the eye pins (Fig. 7f). Whereas previous parts could be built directly by rapid prototyping machines, the membrane (Fig. 7f) needed to be built in several steps due to its complex design criteria, which are the following: • • • •
the membrane needs to be flexible; the stiff eye pins need to be firmly attached; precisely positioned to the membrane; its wall thickness needs to be constant to avoid irregular wall deformation.
Although it is not straightforward to take into account all of these design criteria, we succeeded to build this membrane using the following strategy. In order to build such a membrane, we started to produce a silicone mould for Vacuum Casting. The mould was formed
Fig. 8. The practical realization of (b) the flexible bars (a) connecting the membrane and the supporting frame by the use of the eye pins and (c) modular discs.
B. Van Der Smissen et al. / IRBM 34 (2013) 226–234
231
Fig. 9. The practical realization of (a) the valve housing is depicted in this figure. It contains (d) the valve plane, (c) the mitral entrance channel, (e) the aortic exit channel, and finally (b and g) two mechanical valves. Leakage between valves and valve plane is prevented by (f) O-rings, which are pressed by (c and e) compression nuts.
starting from a positive part that was built from epoxy using the SLA technique. The positive part has the same geometry as the final membrane with the exception of 31 additional bulges. One bulge correspond with the projected geometry of the upper site of an eye pin. This allows us to position all the eye pins at once and at the right position, which is firstly, perpendicularly oriented to the membrane surface and secondly, at the right position to represent the corners of the 16 segments that are often used in clinical echocardiography to identify LV pathologies. In order to form the final membrane, we worked out the following procedure. After the geometry of the positive part was designed in CAD and converted to an STL format, it was built in epoxy material using the SLA technique. Then, the positive part was placed in a box and immersed by liquid silicone. After polymerization, the liquid silicone transformed to a silicone block, which was then opened with a scalpel. Next, the positive part was taken out of the silicone block and left a negative shape of the positive part in the silicone block. At that moment, the silicone block functioned as a mould and the 2component liquid polyurethane was vacuum casted. However, before the actual casting was performed, the 31 eye pins (built by the SLS technique) were placed inside the silicone mould as insert parts. After polymerization of the polyurethane, the final thin-walled membrane including integrated eye pins was formed. Thanks to this procedure, we could meet the complete set of the design criteria to form this complex membrane in three ways. First, a correct position of the eye pins was realized according to the 16 segments and the perpendicular orientation to the membrane’s surface. Secondly, the stiff eye pins are tightly connected to the flexible membrane, because the liquid (unpolymerized) polyurethane encloses the perforated flange of
the eye pin during casting, assuring a tight connection with the membrane after polymerization. Thirdly, the thickness of the membrane is consistently kept constant (at 3 mm) since the centering pin equally distances the inner and outer part of the silicone mould during casting. In this way, a membrane was realized according to all of the design criteria and therefore ready for experiments.
Fig. 10. Once (a) our model was built, we activated it using (f) a piston pump and connected it to a lumped hydraulic windkessel model of the vascular system, including (b) compliance, (c) resistance, and (d) preload. After that, ventricular pressure (pventr ) and aortic flow (Qaorta ) were measured. In order to create a closed circuit, (e) an overflow and a circulation pump were added.
232
B. Van Der Smissen et al. / IRBM 34 (2013) 226–234
Fig. 11. Compared with physiological data (dashed line) of flow and pressure, the experimental results (solid line) show that the model is able to generate realistic data in terms of end-systolic pressure (120 mmHg) and heart rate (57 BPM). Due to the too high after load resistance, the aortic flow in our model (solid line with mean aortic flow of 1.6 L/min) was low compared to human physiological data at rest (dashed line with mean aortic flow of 5 to 6 L/min [12]).
7. Experiments and preliminary results After the model was built, it was actuated (Fig. 10a) using a piston pump (Harvard pump, Harvard Apparatus, Dover, MA) (Fig. 10f) and connected to a lumped hydraulic windkessel model of the vascular system including compliance (Fig. 10b) and resistance (Fig. 10c). We measured (data acquisition using Labview 7, National Instruments, Austin, TX USA) ventricular pressure (disposable transducer DTX/PLUS, Becton Dickinson Critical Care Systems) and aortic flow (Tubing Flow Sensor ME-PXL, Transonic Systems Inc.). Our experimental results (Fig. 11) showed that this model is able to generate fairly realistic data in terms of end-systolic (120 mmHg versus 120 mmHg [12]) and heart rate (57 BPM versus 60 BPM [12]). Due to the too high afterload resistance, the aortic flow in our model was low compared to (human) physiological data (1.6 L/min versus 5 to 6 L/min [12]). We also observed a few considerable differences compared to human data, such as negative values of pressure and flow which are presumably attributable to the sucking property of the piston pump and the fact that the encasing is rigid and filled with non-compressible liquids, creating
a system with little to no compliance to dampen pressure peaks. In order to visualize wall deformation patterns, we made use of a slightly different setup in terms of actuation, because the semi-transparent outer shell did not allow us to clearly observe the deformation patterns. Better optical results were obtained by removing the outer shell. To obtain this, the initial configuration of the setup needed some modifications. Whereas the actuation pump was connected to the external shell in the previous setup, the actuation in this setup is obtained by connecting the actuation pump directly to the LV cavity via the mitral entrance channel. Instead of mechanical valves in the previous configuration, the valve ports are now at mitral site open and at aortic site closed during whole the heart cycle. Since wall deformation is the result of the transmembranal pressure difference, the direction in which the pressure difference is applied does neither affect the final wall deformation, nor influence the controlling function of the flexible bars. For that reason, we can assume that this modified configuration leads to the same deformation, and for that reason we opted this modified configuration of setup to obtain an improved visualization of the wall deformation. Wall deformation patterns nicely
Fig. 12. Wall deformation patterns of our model were clearly visualized in terms of circumferential and longitudinal contraction. This figure shows also (a) the end-systolic (solid line) and (b) end-diastolic wall deformation (dotted line).
B. Van Der Smissen et al. / IRBM 34 (2013) 226–234
demonstrated circumferential as well as longitudinal contraction (Fig. 12). 8. Discussion and conclusion This paper describes the design and development of an inventive hydraulic LV simulator, which achieves more realistic wall deformations than the existing thin-walled membrane models. This was achieved by a thick-wall model in which control can be exerted on the wall using a set of flexible bars, which are distributed evenly over the membrane surface. Our experimental results show that our model is capable of generating fairly realistic data in terms of pressure and wall deformation patterns. The model displays circumferential and longitudinal contraction, confirming the assumption that the LV wall deformation can be controlled using our flexible bar concept. The controlled wall behaviour in our model is a step forward compared to the unpredictable squeezing of thin-walled models, and should contribute to experimental (hydraulic bench) research on LV wall deformation and its interaction with blood flow. One could argue that similar fairly realistic wall deformation patterns may also be realized in the existing models by using an equilibrium volume, which is smaller than the systolic volume. However, this method is rarely applied to date, most possibly because this would result in a dramatic shortening of the membrane’s lifetime. Our present model overcomes this problem and realizes realistic wall deformation even if the equilibrium volume is larger than the systolic volume, whilst providing a longer lifecycle for the membrane. 9. Limitations and future directions In this study, we tested one single prototype of the novel simulator as a concept of proof, and we did not alter material properties to study the impact of these changes. Similarly, functional measurements were done using a simplified after load setup, which was not optimized to mimic human arterial system properties. The pressure and flow measurements were not performed simultaneously with the visualizations of the wall deformation because the external shell was too opaque. In order to make the external shell optically transparent, the shell surface should be smoothened by a polishing treatment after the rapid prototyping techniques building process. The thin membrane (Fig. 3a), which is the inner layer of the ventricular part, is modelled as a truncated ellipsoid, which is a simplification of the real geometry. A more realistic LV geometry could be obtained via segmentation and reconstruction of medical images [28,29] in order to fabricate patient-specific LV models. Finally, even though our conception of the three-layered thick-wall shows some analogies to the human three-layer structure of endocardium, myocardium, and epicardium, they remain inevitably an oversimplification of the heart’s natural anatomy. In the future, heart simulators with naturally shaped LV models will most possibly play an important role in both clinical and research settings. First of all, the upcoming technologies to build patient-specific experimental models starting from medical images, bears promise that complex interventions can be tested
233
in vitro before these will be ultimately applied to the patient. The modular composition of our experimental setup and LV model allows a smooth integration of these new patient-specific LV models. As such, this model opens possibilities providing a simulation platform, which may enhance such clinical practices in the future. Our model demonstrates the wide potential of RP techniques for building experimental hydraulic LV models and moreover presents one of the first LV models in which left ventricular wall deformation can be controlled. Moreover, the current design allows the local stiffening of the ventricular wall. This provides interesting test cases to mimic the wall dynamics in pathological heart conditions such as myocardial infarct. Local stiffening of the LV wall could be obtained by replacing certain flexible bars with stiff bars. Although the wall deformation is controlled in this new model, further work can be done to achieve even more realistic wall deformation patterns by implementing also the torsion motion. A possible idea to achieve this torsion motion is the following. In the present model, the orientation of the flexible bars is perpendicularly oriented to the membrane surface. Alternatively, if the flexible bars are oriented oblique (for example 45◦ ) to the membrane surface, we suppose that this will lead to a wringing effect of the LV wall and therefore, result in the so-called ‘torsion motion’. Including torsion motion in a heart simulator would open a new window of research opportunities for a better understanding of ventricular dysfunction and an improving management of heart failure patients. Acknowledgements The authors would like to acknowledge the collaboration of the Rapid Prototyping team (Rapid Prototyping & Tooling, University College Ghent) during the realization of the model and the technical assistance of Jurgen Deviche (IBiTech BioMMeda Ghent University) during the experiments. References [1] Massie BM, Shah NB. Evolving trends in the epidemiologic factors of heart failure: rationale for preventive strategies and comprehensive disease management. Am Heart J 1997;133(6):703–12. [2] Verdonck P, et al. Computer-controlled in vitro model of the human left heart. Med Biol Eng Comput 1992;30(6):656–9. [3] Verdonck PR, Van Nooten GJ, Van Belleghem Y. Pulse duplicator hydrodynamics of four different bileaflet valves in the mitral position. Cardiovasc Surg 1997;5(6):593–603. [4] Dumont K, et al. Omnicarbon 21 mm aortic valve prosthesis: in vitro hydrodynamic and echo Doppler study. Int J Artif Organs 2002;25(8):783–90. [5] Scotten LN, Walker DK. New laboratory technique measures projected dynamic area of prosthetic heart valves. J Heart Valve Dis 2004;13(1):120–32 [discussion p. 132–3]. [6] Kaminsky R, et al. Flow visualization through two types of aortic prosthetic heart valves using stereoscopic high-speed particle image velocimetry. Artif Organs 2007;31(12):869–79. [7] Kheradvar A, Gharib M. On mitral valve dynamics and its connection to early diastolic flow. Ann Biomed Eng 2009;37(1):1–13. [8] Wang QA, et al. Hydrodynamic evaluation of a minimally invasive heart valve in an isolated aortic root using a modified in vitro model. J Med Devices Transac Asme 2009;3(1):011002.1–6.
234
B. Van Der Smissen et al. / IRBM 34 (2013) 226–234
[9] Rahmani B, et al. Manufacturing and hydrodynamic assessment of a novel aortic valve made of a new nanocomposite polymer. J Biomech 2012;45(7):1205–11. [10] Vuki´cevi´c M, et al. Experimental study of an asymmetric heart valve prototype. Eur J Mech B Fluids 2012;35(0):54–60. [11] De Mey S, et al. Assesment of LV diastolic filling using color M-mode Doppler echocardiography: validation in new hydoraulic model. Biomech Model Mechanobiol 2004;2(3):127–38. [12] Holt BD, Walsch RA. Normal physiology of the cardiovascular system. In: Fuster V, Walsch R, Harrington R, editors. Hurst’s the Heart. New York: McGraw-Hill Professional; 2010. p. 94–117. [13] Malouf JF, et al. Functional anatomy of the heart. In: Fuster V, Walsch R, Harrington R, editors. Hurst’s the Heart. New York: McGraw-Hill Professional; 2010. p. 63–93. [14] Sengupta PP, et al. Left ventricular form and function revisited: applied translational science to cardiovascular ultrasound imaging. J Am Soc Echocardiogr 2007;20(5):539–51. [15] Han HC. An echocardiogram-based 16-segment model for predicting left ventricular ejection fraction improvement. J Theor Biol 2004;228(1):7–15. [16] Yan X, Gu P. A review of rapid prototyping technologies and systems. Comput Aided Des 1996;28(4):307–18. [17] Lantada AD, Morgado PL. Rapid prototyping for biomedical engineering: current capabilities and challenges. In: Yarmush ML, Duncan JS, Gray ML, editors. Annual Review of Biomedical Engineering, vol. 14. Annual Reviews; 2012. p. 73–96. [18] Webb PA. A review of rapid prototyping (RP) techniques in the medical and biomedical sector. J Med Eng Technol 2000;24(4):149–53. [19] Melchels FPW, Feijen J, Grijpma DW. A review on stereolithography and its applications in biomedical engineering. Biomaterials 2010;31(24):6121–30.
[20] Kruth JP, Leu MC, Nakagawa T. Progress in additive manufacturing and rapid prototyping. Cirp Annals 1998 – Manufacturing Technology. In: Alberti N, et al., editors. Annals of the International Institution for Production Engineering Research, vol. 47/2/1998. Elsevier; 1998. p. 525–40. [21] Mankovich NJ, et al. Surgical planning using 3-dimensional imaging and computer modeling. Otolaryngol Clin North Am 1994;27(5): 875–89. [22] Kumar S. Selective laser sintering: a qualitative and objective approach. J Miner Metal Mater Soc 2003;55(10):43–7. [23] Kruth JP, et al. Lasers and materials in selective laser sintering. Assembly Automation 2003;23(4):357–71. [24] O’Neill W, et al. Investigation on multi-layer direct metal laser sintering of 316L stainless steel powder beds. Cirp Annals 1999. Manufacturing Technol 1999;48:151–4. [25] Niu HJ, Chang ITH. Selective laser sintering of gas atomized M2 high speed steel powder. J Mater Sci 2000;35(1):31–8. [26] Kandis M, Bergman TL. A simulation-based correlation of the density and thermal conductivity of objects produced by laser sintering of polymer powders. J Manufacturing Sci Eng Transac Asme 2000;122(3): 439–44. [27] Ho HCH, Gibson I, Cheung WL. Effects of energy density on morphology and properties of selective laser sintered polycarbonate. J Mater Process Technol 1999;90:204–10. [28] Fleureau J, et al. 3D multi-object segmentation of cardiac MSCT imaging by using a multi-agent approach. IRBM 2009;30(3): 104–13. [29] Constantinides C, et al. Development and evaluation of an algorithm for the automated segmentation of the left and right ventricles on cine MRI. IRBM 2009;30(4):188–91.
www.sciencedirect.com IRBM 34 (2013) 226–234
Original article
Modelling the left ventricle using rapid prototyping techniques B. Van Der Smissen a,b,∗ , T. Claessens a,b , P. Verdonck b , P. Van Ransbeeck c , P. Segers b a
University College Ghent, BioMech, 9000 Ghent, Belgium University College Ghent, IBiTech BioMMeda, 9000 Ghent, Belgium c University College Ghent, Department of Mechatronics, 9000 Ghent, Belgium b
Received 16 October 2012; received in revised form 28 March 2013; accepted 1st April 2013 Available online 13 May 2013
Abstract Biomechanical research of left ventricular function involves the assessment and understanding of both ventricular wall mechanics and deformation and intraventricular flow patterns, as well as how they interact. Experimental research using hydraulic bench models should therefore aim for an as realistic as possible simulation of both. In previous experimental investigations, wall deformation was studied by means of thin-walled passive experimental models, consisting of a silicone membrane in a closed box, which is squeezed passively by an externally connected piston pump. Although the pump function of these models has already been well established, the membrane deformation remains unpredictable and the effect of muscle contraction – and hence natural wall deformation – cannot be simulated. In this study, we propose a new design of an experimental hydraulic left ventricular model in which left ventricular wall deformation can be controlled. We built this model by a combination of rapid prototyping techniques and tested it to demonstrate its wall deformation and pump function. Our experiments show that circumferential and longitudinal contraction can be attained and that this model can generate fairly normal values of pressure and flow. © 2013 Elsevier Masson SAS. All rights reserved.
1. Introduction Heart failure is an increasingly important health problem in most developed countries [1]. An important focus in biomechanical research of congestive heart failure is the role of the deformation of the ventricular wall in the interaction with blood flow. Therefore, research using experimental models should aim to a realistic imitation of both. In previous experimental investigations, wall deformation was studied by means of thin-walled passive models, consisting of a silicon membrane in a closed box that is passively squeezed by an externally connected piston pump. Experimental models relying upon this deformation mechanism have been used to study the influence of the different determinants of left heart performance on transmitral flow [2], to test the hydrodynamic performance of heart valves [3–10] or to investigate left ventricular diastolic filling using colour M-mode Doppler echocardiography [11]. Although the pump function of these models is well established, the membrane deformation
remains unpredictable and the effect of muscle contraction – and hence natural wall deformation – cannot be simulated. 2. Objectives In this study, we propose a new design of an experimental hydraulic left ventricular model in which left ventricular wall deformation can be controlled. First, we have built this model by a combination of rapid prototyping techniques and after that tested it to demonstrate its wall deformation and pump function by measuring aortic flow and ventricular pressure. The model, which is described in this article, is the result of an optimization process in which we gradually improved our previous designs based on different design criteria (e.g. practical feasibility). However, in this article, we exclusively describe the design and composition of the final model that was also physically realized. 3. Physiological background
∗
Corresponding author. University College Ghent, BioMech, 9000 Ghent, Belgium. E-mail address: [email protected] (B. Van Der Smissen). 1959-0318/$ – see front matter © 2013 Elsevier Masson SAS. All rights reserved. http://dx.doi.org/10.1016/j.irbm.2013.04.001
The heart is a four-chambered, muscular organ that continuously pumps blood through the body’s extensive network of arteries and veins. The left ventricle (LV) can be considered
B. Van Der Smissen et al. / IRBM 34 (2013) 226–234
227
as its most important and powerful pumping chamber, and is most frequently studied [12]. The LV wall consists of three layers: the epicardium, the myocardium, and the endocardium. The endocardium is the thin inner layer of endothelium while the epicardium is an external visceral layer covering the heart. The myocardium, the middle layer, forms the bulk of the LV and is responsible for the actual contraction [13]. It consists of interweaving bundles of cardiac muscle fibres embedded in connective tissue structure and spirally arranged around the circumference of the heart [13,14]. As a result of this anatomical architecture and of the timing and sequence of electrical excitation, during contraction the LV deforms as follows: (a) the diameter of the LV decreases while (b) the base-apex distance decreases (c) in a rotating manner [14]. As such, global ventricular wall deformation can be basically described in terms of (a) circumferential contraction, (b) longitudinal contraction, and (c) a torsion motion [14].
4. Contraction principle Up till now, experimental left ventricular models have been primarily used to test the hydrodynamic performance of heart valves [3–10] and, to a lesser extent, to study transmitral flow [2] or left ventricular diastolic filling [11]. Given these specific foci of research, not much attention has been directed towards mimicking the complex wall deformation patterns. In all referred models, the ventricle has been modelled as a passive thin-walled silicone membrane within a closed chamber [3–11]. By using a piston pump, water or pressurized air is applied within the closed box and the internal membrane compresses which consequently results in LV volume reduction. Although it is well demonstrated that these passive models provide adequate pump function, the unpredictability of the membrane deformation does not allow simulating the effect of muscle contraction and hence natural wall deformation. In contrast to this thin-walled conception of the LV, we conceived the LV as a thick-walled hydraulic model composed of three layers. First is an inner thin-walled membrane, second is a middle layer that includes flexible bars, which we assume to control the wall deformation and third is an exterior shell. The contraction principle is as following. During the systolic phase, a piston pump pushes a certain amount of fluid into the region with flexible bars, compressing the internal membrane and stretching the flexible bars. In this way, the thin membrane is supported during the whole heart cycle and does not randomly deform as was the case with previous models. This results in wall deformation that is controlled by the flexible bars during systole while LV volume is decreasing. Likewise, during the diastolic phase, fluid is sucked from the region with flexible bars and the wall deforms back to its initial geometry, which also occurs in a controlled manner. These three layers of the ventricular part could be considered as analogues of the human heart. In such an analogy, the thin membrane represents the endocardium, the flexible bars connecting the membrane with a stiff supporting frame represent the myocardium that actually accounts for ventricular contraction,
Fig. 1. Illustrative drawing of a 16-segment left ventricle (LV) model representing the 16 segments, which is often used in echocardiography. The model consists of three slices: (1) the basal and (2) the mid-ventricular slices which are uniformly divided into six segments and (3) the apical slice which is split into four equal segments. Reproduced from Han, 2004 [15].
and the exterior shell covering the latter parts represents the epicardium, which is the external visceral layer of the heart. The degree of control over the deformation of the inner membrane is determined by the tension and number of flexible bars that are arranged over the membrane surface. The higher this number, the more accurate the control of the wall deformation will be. In our model, we provided the possibility to stretch or loosen the flexible bars by adding or removing modular discs in order to control the wall deformation more efficiently (Section 5.1). We have set the number of bars to 31, which represent the corners of a 16-segment model that is often used in clinical echocardiography to define LV pathologies [15]. The 16 segments are performed by dividing the LV model into three slices and each slice into several segments. As such, the basal and the mid-ventricular slices are uniformly divided into six segments and the apical slice into four segments. The configuration of the 16-segment model is illustrated in Fig. 1.
5. Design and composition of the left ventricle model This contraction principle was further optimized in terms of practical feasibility and resulted in the model shown in Fig. 2. In order to facilitate experiments, we modularly designed the model. It consists of two main parts: a ventricular part (Fig. 2b) and a valve housing (Fig. 2a). The ventricular part takes care of the ventricular contraction while the valve housing holds two (mechanical) heart valves and connects the model to an afterload configuration, e.g. a lumped hydraulic windkessel model of the vascular system.
228
B. Van Der Smissen et al. / IRBM 34 (2013) 226–234
Fig. 4. The eye pins are designed with (a) an eye to connect the flexible bars, (d) a perforated flange to assure embedding in a tight connection with the flexible thin membrane, (e) a centering pin to provide (c) a constant membrane thickness of 3 mm and (b) guiding planes to align them perpendicular to the membrane.
Fig. 2. The model consists of two main parts: (a) a valve housing and (b) a ventricular part.
5.1. Design features of the ventricular part
one end to the supporting points in the membrane, the so-called eye pins (Fig. 3g). At the other end, these bars are connected to the supporting frame by modular discs (Fig. 3i). We designed these eye pins (depicted in more detail in Fig. 4) with the following four features:
As depicted in more details in Fig. 3, the ventricular part consists of four components: the thin flexible membrane (Fig. 3a), the flexible bars (Fig. 3b), the stiff supporting frame (Fig. 3c) and the exterior shell (Fig. 3e). First, the thin membrane (Fig. 3a), the inner layer of the ventricular part, is modelled as a truncated ellipsoid with a volume of 130 mL, base-apex length of 85 mm, short axis diameter of 45 mm and with an overall wall thickness of 3 mm. The membrane is made of a 2-component flexible Polyurethane with an E-modulus of 5 MPa. We also provided the membrane with a sealing ring (Fig. 3f) in order to prevent leakage between the region with the flexible bars and the LV cavity. Second, the region with the flexible bars (Fig. 3b), connecting the inner and outer layer of the ventricular part, can be assumed as the middle layer. Importantly, these bars need at the one hand to be flexible enough to allow deformation of the membrane and, at the other hand, to be stiff enough to exert control over the wall deformation. Empirically, we found the following parameters appropriate to account for these requirements in our model, that is a rubber material with an E-modulus of approximately 5 MPa and a cross-section of 4 mm2 . The flexible bars are connected on
• an eye (Fig. 4a) to which a flexible bar finally is connected; • a perforated flange (Fig. 4d) to assure embedding in a tight connection with the flexible thin membrane; • a centering pin (Fig. 4e) to provide a constant membrane thickness; • guiding planes (Fig. 4b) to align them perpendicularly to the membrane (Section 6).
Fig. 3. The ventricular part consists of (a) a thin membrane including (f) a sealing ring, (b) flexible bars, (c) a supporting frame, (d) an O-ring, (e) an external shell, (g) eye pins, (h) supporting points, (i) modular discs and (j) connection to the piston pump.
Fig. 5. The modular discs are designed with (d) a hole and (a) a cut-out in which a transverse spill (Fig. 8c) fits to allow easy connection using different types of flexible bars, (c) a cut-out to facilitate tightening and loosening of the flexible bars and (b) an adjusting ring to fit them together (e).
The modular discs (Fig. 5) were designed also with several features: • a hole (Fig. 5d) where the flexible bars pass through; • a cut-out (Fig. 5a) in which a transverse spill (Fig. 8c) fits in order to allow easy connection by using different types of flexible bars; • a cut-out (Fig. 5c) to facilitate tightening and loosening of the flexible bars; • an adjusting ring (Fig. 5b) to keep the different modular discs together (Fig. 5e). The third component, the supporting frame (Fig. 3c), is located at the outer side of the middle layer of the ventricular part. It serves not only to keep the supporting points (Fig. 3h) in a fixed position, but it also enables visual inspection of the wall deformation as well as a more practically feasible assembly of the flexible bars. We designed its geometry according to
B. Van Der Smissen et al. / IRBM 34 (2013) 226–234
229
6. Rapid prototyping materials and methods
Fig. 6. The valve housing contains (e) a valve plane, (a) two gates and positions, (d) two mechanical valves. Leakage between valves and valve plane is prevented by (c) O-rings, which are pressed by (b) compression nuts.
several design criteria. First, the general geometry is set with a constant offset (25 mm) to the membrane and it consists of a coarse-meshed structure (beam height and width of 3 and 5 mm respectively). In addition, supporting points are included to realize a perpendicular orientation of the flexible sticks with respect to the membrane surface. As such, the flexible bars have a constant length and therefore ensure uniform control of the LV wall deformation. Finally, the exterior shell (Fig. 3e) functions as the exterior box in which fluid is pumped by a piston pump. Similar to the geometry of the supporting frame, the exterior shell is designed with a constant offset distance from the membrane. We have set an offset of 2 cm with respect to the supporting frame to allow stacking of several modular discs (Fig. 5c) intended for extrastretching of the flexible bars. Another feature of the exterior shell is the pump connection, which is located underneath at apex position (Fig. 3j). Leakage from the middle layer of the ventricular part (region with the flexible bars) towards the outside of the model is prevented using a sealing ring (Fig. 3d), which is located between the exterior shell and the valve housing. 5.2. Design features of the valve housing The valve housing assembly consists of a valve housing part (Fig. 6 and Fig. 9) that positions and seals two mechanical heart valves, one at mitral and the other one at aortic position (Fig. 6d and Fig. 9b and g). Each valve is oriented with an angle of 55◦ with respect to the long axis of the LV. Leakage between the valves and the valve plane is prevented using an O-ring (Fig. 6c and Fig. 9f) that is compressed by a compression nut (M42 with a pitch of 1.5 mm) (Fig. 6b and Fig. 9c and e). The valves are positioned as closely as possible to the valve plane in order to minimize the non-contractile space of the LV cavity. However, the limiting factor hereby is the minimal wall thickness of the middle part of the valve housing. This is the place where the distance between both valves is minimal (we set this distance to 2 mm). In order to position the valves even closer to the valve plane, we shaped the two canals with a constant surface area as following: the shape of the cross-section gradually changed starting from a circular-shape at the level of the valves, up to a semicircular-shape towards the LV cavity.
In the past two decades, rapid prototyping (RP) techniques have substantially changed and improved prototyping manufacturing practices in both industry and research. These techniques allow to quickly fabricate three-dimensional products using Computer Aided Design (CAD) systems [16], producing prototypes in a broad range of materials with outstanding precision, whilst taking only a couple of hours of production. These techniques are now successfully integrated into product development methodologies as they allow to address the market’s requirements in a customized way. Being continuously under development, the range of applications of RP technology now also expands to the biomedical field as it provides innovative support for research plans requiring prototypes [17]. RP technologies can be of assistance in every stage of the development process of novel devices, can help to address various problems arising during development and also supports design [17]. Current examples of rapid prototyping technology applications in the medical field include surgical planning, and the design and development of (patient-specific) prostheses [18], surgical tools, implants, and other biomedical devices [19]. In this study, the creation of the thick-walled LV model was made possible using an innovative combination of three different rapid prototyping techniques (Fig. 7), namely Selective Laser Sintering (SLS), Stereolithography (SLA) and vacuum casting. SLS and SLA, comprising two basic techniques of rapid prototyping (also popularly known as 3D printing) that are both additive manufacturing techniques. These techniques differ fundamentally from traditional manufacturing where an object is shaped by material removal, such as cutting, drilling, and milling. In the case of additive manufacturing, the method to produce shaped parts is by creating or adding solid material layer by layer [20]. The designed geometry of the desired object can be created using 3D drawing computer software or can be derived from scanning data of imaging technologies such as magnetic resonance imaging (MRI) or computed tomography (CT) techniques [21]. This geometrical construction is then virtually sliced into thin layers that correspond to the layer thickness provided by the layer by layer printing process [19]. SLA was introduced in the late 1980s and is one of the most powerful and versatile of all rapid prototyping techniques. It has the highest fabrication accuracy and a wide range of materials that can be processed [19]. The material used for SLA is a liquid photo-curable resin, with an ultraviolet laser emitting photons, which initiate the polymerization of monomers [16]. Based on this principle, solid objects are built in a tank of liquid resin by successively curing thin layers, one on top of the other. SLS is a powder-based layer-process. Laser beams are used as a heat source for scanning and fusing powder materials in predetermined sizes and shapes of layers. After the first layer is scanned, a second layer of loose powder is deposited over it, and the process is repeated from bottom to top until the object is complete [22]. One advantage of SLS is that it can be applied with a wide range of materials [23], such as wax, cermet, ceramics, nylon/glass composite, metal-polymer powders,
230
B. Van Der Smissen et al. / IRBM 34 (2013) 226–234
Fig. 7. This figure shows the practical realization of (a) our model, (b) the valve housing, (c) an assembly of the supporting frame and membrane without the flexible bars, (d) the external shell, (e) the supporting frame and (f) the membrane containing the embedded eye pins.
metals, alloys or steels [24,25], polymers [26], nylon, and carbonate [27]. As such, materials can be chosen based on their characteristics dependent on the type of process, and therefore, SLS is primarily used to manufacture functional parts. The third technique we used is vacuum casting, a manufacture process often used to produce small series of elastomer products. This process is not an additive manufacturing techniques (such as SLA and SLS), but a casting technique. It starts by placing a two piece silicone mould in a vacuum chamber. Then, the elastomer is mixed and degassed and subsequently poured into the mould. The vacuum is used to remove the entrapped air and to draw the liquid elastomer material into a silicone mould. After the vacuum is released, the mould is removed from the chamber. Finally, the liquid elastomer is cured in an oven and the mould is removed to release the completed casting. This silicone mould can be reused to produce identical copies of the needed elastomer parts. Application of RP techniques to build the LV model: the combination of these three techniques enabled us to build the complete model, which is composed of three different types of materials, such as: • epoxy (by SLA) to build the external shell; • polyamide (by SLS) to fabricate the supporting frame, valve plate and the tinier parts such as the eye pins;
• 2-component polyurethane (by Vacuum Casting) to produce the LV membrane. The external shell (Fig. 7d) was built from epoxy using the SLA technique because this technique allows semi-transparency in combination with solid properties providing enough strength to deal with the pressurized media. The SLS technique, however, was used to build parts that can resist relative high forces such as the modular discs (Fig. 8c), the supporting frame (Fig. 7e), the valve plane (Fig. 9) and the eye pins (Fig. 7f). Whereas previous parts could be built directly by rapid prototyping machines, the membrane (Fig. 7f) needed to be built in several steps due to its complex design criteria, which are the following: • • • •
the membrane needs to be flexible; the stiff eye pins need to be firmly attached; precisely positioned to the membrane; its wall thickness needs to be constant to avoid irregular wall deformation.
Although it is not straightforward to take into account all of these design criteria, we succeeded to build this membrane using the following strategy. In order to build such a membrane, we started to produce a silicone mould for Vacuum Casting. The mould was formed
Fig. 8. The practical realization of (b) the flexible bars (a) connecting the membrane and the supporting frame by the use of the eye pins and (c) modular discs.
B. Van Der Smissen et al. / IRBM 34 (2013) 226–234
231
Fig. 9. The practical realization of (a) the valve housing is depicted in this figure. It contains (d) the valve plane, (c) the mitral entrance channel, (e) the aortic exit channel, and finally (b and g) two mechanical valves. Leakage between valves and valve plane is prevented by (f) O-rings, which are pressed by (c and e) compression nuts.
starting from a positive part that was built from epoxy using the SLA technique. The positive part has the same geometry as the final membrane with the exception of 31 additional bulges. One bulge correspond with the projected geometry of the upper site of an eye pin. This allows us to position all the eye pins at once and at the right position, which is firstly, perpendicularly oriented to the membrane surface and secondly, at the right position to represent the corners of the 16 segments that are often used in clinical echocardiography to identify LV pathologies. In order to form the final membrane, we worked out the following procedure. After the geometry of the positive part was designed in CAD and converted to an STL format, it was built in epoxy material using the SLA technique. Then, the positive part was placed in a box and immersed by liquid silicone. After polymerization, the liquid silicone transformed to a silicone block, which was then opened with a scalpel. Next, the positive part was taken out of the silicone block and left a negative shape of the positive part in the silicone block. At that moment, the silicone block functioned as a mould and the 2component liquid polyurethane was vacuum casted. However, before the actual casting was performed, the 31 eye pins (built by the SLS technique) were placed inside the silicone mould as insert parts. After polymerization of the polyurethane, the final thin-walled membrane including integrated eye pins was formed. Thanks to this procedure, we could meet the complete set of the design criteria to form this complex membrane in three ways. First, a correct position of the eye pins was realized according to the 16 segments and the perpendicular orientation to the membrane’s surface. Secondly, the stiff eye pins are tightly connected to the flexible membrane, because the liquid (unpolymerized) polyurethane encloses the perforated flange of
the eye pin during casting, assuring a tight connection with the membrane after polymerization. Thirdly, the thickness of the membrane is consistently kept constant (at 3 mm) since the centering pin equally distances the inner and outer part of the silicone mould during casting. In this way, a membrane was realized according to all of the design criteria and therefore ready for experiments.
Fig. 10. Once (a) our model was built, we activated it using (f) a piston pump and connected it to a lumped hydraulic windkessel model of the vascular system, including (b) compliance, (c) resistance, and (d) preload. After that, ventricular pressure (pventr ) and aortic flow (Qaorta ) were measured. In order to create a closed circuit, (e) an overflow and a circulation pump were added.
232
B. Van Der Smissen et al. / IRBM 34 (2013) 226–234
Fig. 11. Compared with physiological data (dashed line) of flow and pressure, the experimental results (solid line) show that the model is able to generate realistic data in terms of end-systolic pressure (120 mmHg) and heart rate (57 BPM). Due to the too high after load resistance, the aortic flow in our model (solid line with mean aortic flow of 1.6 L/min) was low compared to human physiological data at rest (dashed line with mean aortic flow of 5 to 6 L/min [12]).
7. Experiments and preliminary results After the model was built, it was actuated (Fig. 10a) using a piston pump (Harvard pump, Harvard Apparatus, Dover, MA) (Fig. 10f) and connected to a lumped hydraulic windkessel model of the vascular system including compliance (Fig. 10b) and resistance (Fig. 10c). We measured (data acquisition using Labview 7, National Instruments, Austin, TX USA) ventricular pressure (disposable transducer DTX/PLUS, Becton Dickinson Critical Care Systems) and aortic flow (Tubing Flow Sensor ME-PXL, Transonic Systems Inc.). Our experimental results (Fig. 11) showed that this model is able to generate fairly realistic data in terms of end-systolic (120 mmHg versus 120 mmHg [12]) and heart rate (57 BPM versus 60 BPM [12]). Due to the too high afterload resistance, the aortic flow in our model was low compared to (human) physiological data (1.6 L/min versus 5 to 6 L/min [12]). We also observed a few considerable differences compared to human data, such as negative values of pressure and flow which are presumably attributable to the sucking property of the piston pump and the fact that the encasing is rigid and filled with non-compressible liquids, creating
a system with little to no compliance to dampen pressure peaks. In order to visualize wall deformation patterns, we made use of a slightly different setup in terms of actuation, because the semi-transparent outer shell did not allow us to clearly observe the deformation patterns. Better optical results were obtained by removing the outer shell. To obtain this, the initial configuration of the setup needed some modifications. Whereas the actuation pump was connected to the external shell in the previous setup, the actuation in this setup is obtained by connecting the actuation pump directly to the LV cavity via the mitral entrance channel. Instead of mechanical valves in the previous configuration, the valve ports are now at mitral site open and at aortic site closed during whole the heart cycle. Since wall deformation is the result of the transmembranal pressure difference, the direction in which the pressure difference is applied does neither affect the final wall deformation, nor influence the controlling function of the flexible bars. For that reason, we can assume that this modified configuration leads to the same deformation, and for that reason we opted this modified configuration of setup to obtain an improved visualization of the wall deformation. Wall deformation patterns nicely
Fig. 12. Wall deformation patterns of our model were clearly visualized in terms of circumferential and longitudinal contraction. This figure shows also (a) the end-systolic (solid line) and (b) end-diastolic wall deformation (dotted line).
B. Van Der Smissen et al. / IRBM 34 (2013) 226–234
demonstrated circumferential as well as longitudinal contraction (Fig. 12). 8. Discussion and conclusion This paper describes the design and development of an inventive hydraulic LV simulator, which achieves more realistic wall deformations than the existing thin-walled membrane models. This was achieved by a thick-wall model in which control can be exerted on the wall using a set of flexible bars, which are distributed evenly over the membrane surface. Our experimental results show that our model is capable of generating fairly realistic data in terms of pressure and wall deformation patterns. The model displays circumferential and longitudinal contraction, confirming the assumption that the LV wall deformation can be controlled using our flexible bar concept. The controlled wall behaviour in our model is a step forward compared to the unpredictable squeezing of thin-walled models, and should contribute to experimental (hydraulic bench) research on LV wall deformation and its interaction with blood flow. One could argue that similar fairly realistic wall deformation patterns may also be realized in the existing models by using an equilibrium volume, which is smaller than the systolic volume. However, this method is rarely applied to date, most possibly because this would result in a dramatic shortening of the membrane’s lifetime. Our present model overcomes this problem and realizes realistic wall deformation even if the equilibrium volume is larger than the systolic volume, whilst providing a longer lifecycle for the membrane. 9. Limitations and future directions In this study, we tested one single prototype of the novel simulator as a concept of proof, and we did not alter material properties to study the impact of these changes. Similarly, functional measurements were done using a simplified after load setup, which was not optimized to mimic human arterial system properties. The pressure and flow measurements were not performed simultaneously with the visualizations of the wall deformation because the external shell was too opaque. In order to make the external shell optically transparent, the shell surface should be smoothened by a polishing treatment after the rapid prototyping techniques building process. The thin membrane (Fig. 3a), which is the inner layer of the ventricular part, is modelled as a truncated ellipsoid, which is a simplification of the real geometry. A more realistic LV geometry could be obtained via segmentation and reconstruction of medical images [28,29] in order to fabricate patient-specific LV models. Finally, even though our conception of the three-layered thick-wall shows some analogies to the human three-layer structure of endocardium, myocardium, and epicardium, they remain inevitably an oversimplification of the heart’s natural anatomy. In the future, heart simulators with naturally shaped LV models will most possibly play an important role in both clinical and research settings. First of all, the upcoming technologies to build patient-specific experimental models starting from medical images, bears promise that complex interventions can be tested
233
in vitro before these will be ultimately applied to the patient. The modular composition of our experimental setup and LV model allows a smooth integration of these new patient-specific LV models. As such, this model opens possibilities providing a simulation platform, which may enhance such clinical practices in the future. Our model demonstrates the wide potential of RP techniques for building experimental hydraulic LV models and moreover presents one of the first LV models in which left ventricular wall deformation can be controlled. Moreover, the current design allows the local stiffening of the ventricular wall. This provides interesting test cases to mimic the wall dynamics in pathological heart conditions such as myocardial infarct. Local stiffening of the LV wall could be obtained by replacing certain flexible bars with stiff bars. Although the wall deformation is controlled in this new model, further work can be done to achieve even more realistic wall deformation patterns by implementing also the torsion motion. A possible idea to achieve this torsion motion is the following. In the present model, the orientation of the flexible bars is perpendicularly oriented to the membrane surface. Alternatively, if the flexible bars are oriented oblique (for example 45◦ ) to the membrane surface, we suppose that this will lead to a wringing effect of the LV wall and therefore, result in the so-called ‘torsion motion’. Including torsion motion in a heart simulator would open a new window of research opportunities for a better understanding of ventricular dysfunction and an improving management of heart failure patients. Acknowledgements The authors would like to acknowledge the collaboration of the Rapid Prototyping team (Rapid Prototyping & Tooling, University College Ghent) during the realization of the model and the technical assistance of Jurgen Deviche (IBiTech BioMMeda Ghent University) during the experiments. References [1] Massie BM, Shah NB. Evolving trends in the epidemiologic factors of heart failure: rationale for preventive strategies and comprehensive disease management. Am Heart J 1997;133(6):703–12. [2] Verdonck P, et al. Computer-controlled in vitro model of the human left heart. Med Biol Eng Comput 1992;30(6):656–9. [3] Verdonck PR, Van Nooten GJ, Van Belleghem Y. Pulse duplicator hydrodynamics of four different bileaflet valves in the mitral position. Cardiovasc Surg 1997;5(6):593–603. [4] Dumont K, et al. Omnicarbon 21 mm aortic valve prosthesis: in vitro hydrodynamic and echo Doppler study. Int J Artif Organs 2002;25(8):783–90. [5] Scotten LN, Walker DK. New laboratory technique measures projected dynamic area of prosthetic heart valves. J Heart Valve Dis 2004;13(1):120–32 [discussion p. 132–3]. [6] Kaminsky R, et al. Flow visualization through two types of aortic prosthetic heart valves using stereoscopic high-speed particle image velocimetry. Artif Organs 2007;31(12):869–79. [7] Kheradvar A, Gharib M. On mitral valve dynamics and its connection to early diastolic flow. Ann Biomed Eng 2009;37(1):1–13. [8] Wang QA, et al. Hydrodynamic evaluation of a minimally invasive heart valve in an isolated aortic root using a modified in vitro model. J Med Devices Transac Asme 2009;3(1):011002.1–6.
234
B. Van Der Smissen et al. / IRBM 34 (2013) 226–234
[9] Rahmani B, et al. Manufacturing and hydrodynamic assessment of a novel aortic valve made of a new nanocomposite polymer. J Biomech 2012;45(7):1205–11. [10] Vuki´cevi´c M, et al. Experimental study of an asymmetric heart valve prototype. Eur J Mech B Fluids 2012;35(0):54–60. [11] De Mey S, et al. Assesment of LV diastolic filling using color M-mode Doppler echocardiography: validation in new hydoraulic model. Biomech Model Mechanobiol 2004;2(3):127–38. [12] Holt BD, Walsch RA. Normal physiology of the cardiovascular system. In: Fuster V, Walsch R, Harrington R, editors. Hurst’s the Heart. New York: McGraw-Hill Professional; 2010. p. 94–117. [13] Malouf JF, et al. Functional anatomy of the heart. In: Fuster V, Walsch R, Harrington R, editors. Hurst’s the Heart. New York: McGraw-Hill Professional; 2010. p. 63–93. [14] Sengupta PP, et al. Left ventricular form and function revisited: applied translational science to cardiovascular ultrasound imaging. J Am Soc Echocardiogr 2007;20(5):539–51. [15] Han HC. An echocardiogram-based 16-segment model for predicting left ventricular ejection fraction improvement. J Theor Biol 2004;228(1):7–15. [16] Yan X, Gu P. A review of rapid prototyping technologies and systems. Comput Aided Des 1996;28(4):307–18. [17] Lantada AD, Morgado PL. Rapid prototyping for biomedical engineering: current capabilities and challenges. In: Yarmush ML, Duncan JS, Gray ML, editors. Annual Review of Biomedical Engineering, vol. 14. Annual Reviews; 2012. p. 73–96. [18] Webb PA. A review of rapid prototyping (RP) techniques in the medical and biomedical sector. J Med Eng Technol 2000;24(4):149–53. [19] Melchels FPW, Feijen J, Grijpma DW. A review on stereolithography and its applications in biomedical engineering. Biomaterials 2010;31(24):6121–30.
[20] Kruth JP, Leu MC, Nakagawa T. Progress in additive manufacturing and rapid prototyping. Cirp Annals 1998 – Manufacturing Technology. In: Alberti N, et al., editors. Annals of the International Institution for Production Engineering Research, vol. 47/2/1998. Elsevier; 1998. p. 525–40. [21] Mankovich NJ, et al. Surgical planning using 3-dimensional imaging and computer modeling. Otolaryngol Clin North Am 1994;27(5): 875–89. [22] Kumar S. Selective laser sintering: a qualitative and objective approach. J Miner Metal Mater Soc 2003;55(10):43–7. [23] Kruth JP, et al. Lasers and materials in selective laser sintering. Assembly Automation 2003;23(4):357–71. [24] O’Neill W, et al. Investigation on multi-layer direct metal laser sintering of 316L stainless steel powder beds. Cirp Annals 1999. Manufacturing Technol 1999;48:151–4. [25] Niu HJ, Chang ITH. Selective laser sintering of gas atomized M2 high speed steel powder. J Mater Sci 2000;35(1):31–8. [26] Kandis M, Bergman TL. A simulation-based correlation of the density and thermal conductivity of objects produced by laser sintering of polymer powders. J Manufacturing Sci Eng Transac Asme 2000;122(3): 439–44. [27] Ho HCH, Gibson I, Cheung WL. Effects of energy density on morphology and properties of selective laser sintered polycarbonate. J Mater Process Technol 1999;90:204–10. [28] Fleureau J, et al. 3D multi-object segmentation of cardiac MSCT imaging by using a multi-agent approach. IRBM 2009;30(3): 104–13. [29] Constantinides C, et al. Development and evaluation of an algorithm for the automated segmentation of the left and right ventricles on cine MRI. IRBM 2009;30(4):188–91.