Rotary blood pumps (RBPs) are used widely for support of failing hearts, both inside and outside the operating room. Their use as ventricular assist devices gains importance over positive displacement pumps because of their compact size, noise-free operation, and reduced number of moving components. Still, there are some concerns about the nonphysiologic hemodynamics they produce, specifically related to diminished arterial pulse. Also, in the event of pump failure, the absence of valves creates an additional risk of severe regurgitant flow, imposing an extra load on the impaired heart. Recent developments of controllers capable of operating RBPs in pulsatile control modes may address some of the former concerns but they have yet to be widely adopted in the clinical settings, in part because of the lack of characterization of the influence of pulsatile mode upon overall hemodynamics.
The Medos Deltastream (Medos Medizintechnik GmbH, Stolberg, Germany) is a rotary blood pump with a diagonal flow path that is intended for use in miniaturized cardiopulmonary bypass circuits or extracorporeal membrane oxygenation systems.1,2 The device is also applicable for short-term ventricular assist, e.g., to support patients with postcardiotomy heart failure. Despite 5 years of clinical experience and numerous animal trials, detailed investigation of the heart-device interaction has yet to be performed. Previous studies were mainly focused on design optimization and the safety and durability of the device, rather than its unloading capacities.3–5 The pulsatile mode of operation of this device is produced by sinusoidal modulation of the rotational speed. Limited hemodynamic studies in vitro5 have demonstrated that the resulting pressure and flow waveforms do not fulfill the criteria for true pulsatile flow as stated in the literature.6 Nevertheless, the pulsatile pattern may have an important impact on the loading conditions of the heart, and it may have potential to support the failing heart in a superior way compared with continuous-mode operation.
This study investigates the interaction of the Deltastream device with an in vitro heart simulator in both continuous and pulsatile mode. Special attention is given to the load that is imposed on the heart by the device, but no autonomous nervous feedback mechanisms are simulated and thus only the pure hydraulic interaction was assessed.
Materials and Methods
An in vitro setup was constructed that consisted of a pneumatically driven heart simulator, an arterial afterload section, and the Medos Deltastream device (Figure 1). The heart simulator consists of a left atrium and a left ventricle (LV), made out of silicone rubber and contained in an acrylic housing filled with water. Details of the system were described earlier.7 For this study, the ventricle was extended with a silicone tube at the apex to allow apical cannulation. The afterload section, also described previously,8 consists of an air chamber (windkessel), a resistor with compressible foam, and a downstream overflow reservoir. These components respectively mimic the aortic compliance, arteriolar resistance, and venous pressure. The Deltastream pump was connected between the heart simulator and the windkessel so as to simulate apical to ascending aorta cannulation. The pump inflow cannula was 170 mm long with an inner diameter of 10 mm, and the outflow cannula consisted of 190 mm of the same tubing, and an additional 70 mm of ½″ tubing to accommodate two separate flow probes for pump control and data acquisition. The test fluid was a 60/40 V% water-glycerin mixture with a kinematic viscosity of 3.65 mm2/s at an average temperature of 26°C.
The standard driver of the heart simulator is a computer-controlled pneumatic actuation system where a proportional valve generates a programmed pressure waveform based on proportional-integral feedback of left ventricular pressure. However, such a feedback control prevents unloading of the ventricle, where lower pressures than the predetermined curve are expected. The feedback pressure transducer was, therefore, set to a fixed (threshold) pressure value with a water column. As a consequence, the proportional valve acted as a simple on/off valve that was switched when the programmed pressure waveform crossed the threshold pressure. This resulted in an operation similar to the older drivers of pneumatic assist devices (e.g., Thoratec, Medos), where systole is obtained by opening a valve between a pressurized buffer reservoir and the device. These drivers usually apply a slightly negative pressure through a separate buffer reservoir to facilitate pump filling. Our system, however, was vented to atmospheric pressure.
The Deltastream pump was driven by a laboratory drive console, provided by the manufacturer, that allowed both fixed speed and “pulsatile” operation mode. In fixed speed mode, the rotary speed (rpm) setpoint is prescribed by the user. The pulsatile control mode allowed the user to set the pulse frequency (“pump rate”), outflow pulse pressure, and average flow rate. The software thereby imposed a sinusoidal rotational speed pattern upon the pump to achieve these setpoints.
The console features two pressure sensors for feedback, which were connected to the LV (preload) and the windkessel (afterload). A flow probe—clipped on the pump outflow cannula—was also required. No triggering was available on the Deltastream drive console and, consequently, only asynchronous pulsatile support could be simulated. In addition to pump inlet and outlet pressures and pump flow, the rotational speed and electrical power were logged automatically by the console at approximately 10 Hz. Further data were acquired with the controller console of the heart simulator at a sample frequency of 200 Hz for at least 5 seconds under steady-state conditions. These data included left ventricular pressure (PLV) measured with a hi-fi cathetertip transducer (Millar Instruments Inc., Houston, TX) and aortic pressure (Pao) measured in the windkessel with a DTX+ strain gauge pressure transducer (Ohmeda, Gent, Belgium). Aortic flow (Qao) was measured at the connecting tube between the aortic valve and the windkessel with a H16C probe and HT207 flow meter (Transonic Systems Inc., Ithaca, NY) while the pump flow (Qpump) was sampled at the pump outflow cannula with an H11X flow probe and HT109 flow meter. Additionally, the volume in the left ventricle (VLV) was acquired from an ultrasonic level meter (Superprox SM606, Hyde Park Electronics LLC, Dayton, OH) in the water column that communicates with the ventricle housing for pneumatic actuation.
Synchronization of the data captured by the heart simulator console and the Deltastream drive console was achieved by providing an external clock pulse (approximately 0.2 Hz) to both devices through a separate function generator.
The Deltastream pump was operated according to a protocol that started with a baseline acquisition (BL) where the pump was shut off and the inflow cannula clamped. Next, the clamp was removed to simulate pump failure (0 rpm), which was followed by incrementally increasing the rotational speed from 1000 to 5000 rpm in steps of 1000 rpm. At each of the seven steps in this pump protocol, data acquisition was performed when a steady state was obtained. This pump protocol was repeated for various “contractile” and rhythmic conditions of the heart: The heart’s contractile state (systolic performance) was varied by altering the pressure of the pneumatic buffer reservoir in the actuating system. The pressure level (PL) was set to obtain systolic aortic pressures of 80, 100, and 120 mm Hg at baseline, and these conditions are further indicated as PL80, PL100, and PL120, respectively. PL120 simulated a healthy subject whereas PL100 mimics the hemodynamic state of that subject with mild heart failure and PL80 represents severe heart failure. Within each pressure level, the heart rate (HR) was set consecutively at 50, 100, and 150 beats per minute (bpm), where 100 bpm comes closest to the clinically observed heart rates. The in vitro afterload components were set to obtain a total arterial compliance of 0.9 ml/mm Hg and a total vascular resistance of 1.2 mm Hg·s/ml at PL120 and a heart rate of 100 bpm. The afterload components were not varied during the measurements and thus there were 63 different conditions: 7 pump rates, 3 heart rates, and 3 contractile states that were all combined.
All experiments in pulsatile mode were performed at PL100. Two combinations of heart rate and pump rate (PR) were tested at two different levels of pump flow and pulse pressure to assess the interaction between the pump and the heart. An overview of the four tested combinations is given in Table 1. Data were acquired repeatedly over 10-second spans to observe the changes in flow and pressure patterns over time.
Furthermore, a simulation of synchronous ventricular assist was performed where different trigger settings were analyzed. This was achieved by combining a heart rate of 100 bpm (at PL100) with a pump rate of 95 bpm while data were acquired over 150 consecutive heart beats (HR 100 – PR 95). This was repeated with a PR of 98 bpm where 165 heart beats were acquired (HR 100 – PR 98). As a result of the small difference in frequencies, each heart beat and pump beat had approximately the same period, but for each consecutive beat the onset of the pump “beat” was slightly delayed with respect to the onset of the heart beat. In this way, co-pulsation, counterpulsation, and all intermediate triggering forms were simulated. For each beat, the lag between the onsets of pump and heart beat was calculated and expressed as a percentage of the heart period. The onset of the heart beat was defined as end-diastole, whereas the onset of a pump beat was defined as the point where the rotational speed falls below the average of the sinusoidal speed variation pattern (Figure 2), which relates to the minimal derivative.
The data captured with the driving consoles of the heart simulator and the Deltastream pump were consecutively synchronized, interpolated, and merged to obtain one data file per experimental condition. For the continuous-mode data, an averaged beat was assembled from this file that was used for further calculation of average pressures and flows, hydraulic power (= flow × pressure head), stroke volume (SV), end-diastolic and end-systolic volume (EDV, ESV), and pressure volume area (PVA). The latter parameter is the summation of the area within the left ventricular pressure-volume loop and the area between the end-systolic and end-diastolic pressure-volume relation in the same plane. It has been demonstrated that the PVA is a measure of oxygen consumption of the myocardium in a real heart.9
For the calculation of the pressure-volume relations, a volume intercept (V0) of 135 ml was assumed, based on the results of the pulsatile mode simulations (Figure 6). V0 is the volume at which the ventricle cannot generate any pressure and it is therefore an intrinsic property of a ventricle and its contraction mechanism. In our in vitro model, however, V0 is related to the dead volume in the rubber ventricle, and thus PVA and other volume-related parameters should not be compared blindly to clinically observed values. Nonetheless, these parameters serve well as a means to compare the different interactions simulated in this experiment.
The pressure, flow, volume, and power parameters that were derived for the continuous mode simulations were also calculated for the pulsatile mode simulations, but based on individual beats instead of an averaged beat.
An example of baseline pressure and flow data is shown in panel A of Figure 5 for a heart rate of 100 bpm and a pressure level of PL100. The results of the generated power and the resulting flows are displayed in Figure 3 for the different pressure levels and rotational speeds for a heart rate of 100 bpm. The generated heart power was observed to increase with increasing pressure level, while the pump power was found to be mainly sensitive to pump speed. As a result, the total power also increased with pump speed. The power that was generated by the heart at baseline is higher than any other. At a speed of 5000 rpm, the cardiac power was calculated to be negative, indicating that the heart “absorbs” power. This observation was found to be concomitant with aortic regurgitation.
The total generated flow remained constant (mean ± SD: 4.8 ± 0.3 l/min), while the balance between pump flow and aortic flow shifted in favor of the pump flow at higher rotational speeds.
A simulation of pump failure (pump speed = 0 rpm) resulted in regurgitation through the assist device of up to 2 l/min, depending on the pressure level of the heart. At a heart rate of 150 bpm, this regurgitant flow increased to 3 l/min. Because the generated cardiac output also increased, net total flow was 4.7 ± 0.3 l/min—equivalent to the previous conditions. A heart rate of 50 bpm had an opposite effect, yielding a regurgitant flow of only 1 l/min and reduced cardiac output, resulting in a net aortic flow of 4.4 ± 0.3 l/min. As a result of pump failure (0 rpm), the mean aortic pressure (MAP) dropped from 80 mm Hg at baseline (for PL100) to 43 mm Hg (Figure 4). Increasing pump speed augmented the pressure gradually to a value of 147 mm Hg at 5000 rpm. A heart rate of 150 bpm results in a MAP of 178 mm Hg for the same condition. A similar rising trend was observed for the PVA, where the highest values are reached at a heart rate of 50 bpm: 10,445 mm Hg·ml at baseline (PL100), compared with 8,177 mm Hg·ml for the baseline value at 100 bpm.
Pump failure resulted in an increase in stroke volume of 6 ml compared with BL (PL100, HR 100 bpm), and it takes a rotational pump speed of 3000 rpm to get the SV back to the baseline level. At a speed of 5000 rpm, the stroke volume was found to decrease to 36 ml—37% below baseline. For the same pressure and heart rate conditions, end-diastolic volume increased 10 ml from baseline if pump failure was simulated and it could be decreased 5 ml from baseline by setting a pump speed of 5000 rpm. Both SV and EDV were found to be highly dependent on the heart rate: while baseline SV and EDV at 100 bpm (PL100) are 57 and 235 ml, respectively, these values are 104 and 265 ml at 50 bpm, and 39 and 254 ml at 150 bpm. In fact, at 150 bpm the EDV was not reduced with increasing pump speed, but gradually rose to a value of 262 ml at 5000 rpm.
A few examples of the heart-device interaction are shown in Figure 5. Panels B and C show two samples of the combination of a heart rate of 100 bpm with a pump rate of 50 bpm. It can be observed that a pattern of two different alternating beats appears, which is most distinct in panel B. In panel C, which was acquired approximately 3 minutes later, the pattern is also present, but the two beats are very similar and, consequently, the aortic pressure waveform looks almost normal. Although the left ventricular pressure is generated by the heart at a rate of 100 bpm, the flow waveforms of panel C indicate that the ventricle is emptied alternately via the aortic valve and the pump. At times when the pump “ejects,” outflow through the aortic valve is completely suppressed. The mean pump flow for this combination was set at 3.5 l/min on the drive console; however, the independent flow measurement indicated 4.3 l/min. Pulse pressure generated by the pump was set at 40 mm Hg on the console, resulting in a measured level of 39 mm Hg in the data of panel B, whereas only 28 mm Hg was reached in the data of panel C.
Panel D shows a sample in which heart and pump rate were set at 80 bpm and 60 bpm respectively (Combination 3 in Table 1). The mean pump flow was set at 2 l/min and yielded 2.4 l/min, while the pulse pressure was set at 20 mm Hg on the console, yet yielded 86 mm Hg if defined as the difference between maximum and minimum aortic pressure over the entire sample. Panel D further shows how a pattern of four consecutive heart beats is formed. This can best be observed in the left ventricular pressure, where each beat differs distinctively from the previous in waveform and amplitude. The aortic pressure also demonstrates a recurring pattern, which deviates very much from a normal aortic pressure pattern and exhibits various waveform shapes and pressure levels. The aortic flow is characterized in the same way, while the pump flow seems to display a pattern that consists of three beats in the same time span as the four beats of the pressure patterns.
The simulation of synchronous ventricular assist at different trigger settings (phase shifts), by combining a heart rate of 100 bpm with a pump rate of 98 bpm, results in a sinusoidal varying pattern in LV volume and pressure, as displayed in Figure 6 (top panel). This translates to a gradual shape change of the pressure-volume loops, as displayed in the bottom panel of Figure 6. The loop with the highest stroke volume yielded the lowest systolic ventricular pressure and vice versa. The loop with the highest stroke volume was attained when a phase shift of 38% was present. According to the phase shift definition given above, this is consistent with a “co-pulsation” mode, in which the rotational speed decreases steeply at end-systole. The loop with the highest systolic ventricular pressure and lowest stroke volume occurred at a phase shift of 83%, which can be considered as counter-pulsation. The top panel of Figure 6 also displays the effect of the phase shifts on aortic pressure, where the pulse pressure varies between 47 mm Hg and 28 mm Hg.
The varying phase shifts induced changes in loading conditions of the LV and consequently an end-systolic pressure-volume relationship (ESPVR) can be determined. The ESPVR was assumed linear and an iteration method10 was used to estimate V0 (135 ml) which was later used in the calculation of PVA.
An overview of averaged parameters is given in Figure 7 for the whole range of phase shifts. It is found for the HR 100 – PR 98 combination that the minimum MAP and PVA occur approximately at a 30% phase shift, while SV and EDV exhibit their minimum at 83% and 64%, respectively. MAP varies between 74 and 133 mm Hg and the stroke volume between 44 and 64 ml. The mean pump flow, however, remained constant at 2.4 ± 0.1 l/min over the whole phase shift range, whereas no trend could be found for the mean aortic flow, which yielded an overall average of 2.7 ± 0.7 l/min.
Interpolation of the electrical power data from the continuous speed experiment (PL100, HR: 100 bpm), implies that the Deltastream pump would consume 3451 mW power to maintain a constant pump flow of 2.4 l/min. In contrast, the average power consumption during the pulsatile HR 100 – PR 98 experiment was 4212 mW. However, power and pump efficiency also varied according to the same sinusoidal pattern as MAP, but in the ranges 2153–6983 mW and −3% to 16%, respectively.
The use of a rotary blood pump at a low continuous rotational speed can lead to regurgitation of fluid from the aorta to the LV through the pump. This, however, is compensated by a higher cardiac output in order to keep the total flow constant. A similar phenomenon was previously observed in healthy animals,11,12 but here it is clear that there are no biological compensation mechanisms involved. A possible explanation is that the ventricle fills more efficiently during regurgitation, and that the pneumatic actuation system can subsequently eject more fluid. A similar Starling-like response has been reported earlier in a pneumatically actuated mock loop.13
Figure 3 indicates that the net flow through the pump at low rotational speed is negative, which means that the integral negative pump flow during heart diastole is larger than the integral positive pump flow that is generated during systole. This is due to three reasons: the longer duration of diastole, the higher (negative) pressure head during diastole, and the functioning of the aortic valve. The last two reasons are closely related: the operation of the valve restricts the regurgitant flow to pass only through the pump, while forward flow is split between pump and aorta. Opening of the valve also makes that ventricular and aortic pressure equilibrate, while the closed valve can bear a high pressure difference. However, the presence of the regurgitant flow path through the pump also alters the pressures, which can be determined from panel A in Figure 4. The mean aortic pressure is low at a pump speed of 0 rpm because it communicates with the zero ventricular pressure during diastole. At high rotational speeds, the added pump energy is established as a pressure rise since the total flow stays constant. The MAP becomes extremely high because the pump continues to generate flow throughout the cardiac cycle, and because of the nonlinearity of the systemic resistor in combination with the more continuous flow pattern.
In accordance with aortic valve insufficiency, regurgitation through the pump leads to a volume overload of the ventricle. This can be observed in the changes in SV and EDV in Figure 4. Although the relative increase in EDV seems small, it should be stressed that the in vitro ventricle has a low ejection fraction and a large nonfunctional volume, which is comparable to a dilated failing heart. The strong dependency of SV and EDV on the heart rate is also a limitation of our in vitro system, which lacks physiologic reflex mechanisms and which fills at atmospheric pressure. Consequently, a lower heart rate results in more time for filling and ejection, thus yielding a higher EDV and SV.
The combination of a pulsatile rotary blood pump with a pumping LV clearly results in large, unphysiologic variations in cardiovascular pressures and flows and the resulting load on the heart. A similar interaction can be expected for displacement type assist devices, although the waveforms produced by such a pump are more physiological. The pulsatile mode of the Deltastream pump theoretically generates a sinusoidal rotational speed pattern, and thus adds energy to the cardiovascular system in such a way. For characterization of the Deltastream pump and controller, it would be interesting to assess the response of the control algorithm to the observed pressure variations. However, such an analysis could not be performed due to the undersampling of the pump speed on the drive console (10 Hz). Judging from the extreme pressure variations, it can be stated that an appropriate control algorithm is mandatory to compensate for this unphysiologic behavior.
If the combination of the two pulsating “pumps: (i.e., the heart and Deltastream) is compared with harmonic wave transmission theory, two phenomena can be observed. One is interference of waves at equal frequencies, where constructive and destructive interference can occur, dependent on the phase shift between the two waves. If the two waves are in phase (co-pulsation), the signals add up and the amplitude becomes the sum of the two wave amplitudes. If the waves are in counterpulsation, the signals cancel out and the amplitude becomes small. This is similar to the data acquired at a heart rate of 100 bpm and a pump rate of 98 (if the difference in frequency is neglected), where the whole range of phase shifts was observed. This leads to times when there is co-pulsation and thus increase in amplitude, and times when there is a reduced amplitude. It should be taken into account that the pressure waves generated by the heart are a superposition of several harmonic waves and that the resulting interference is therefore more complex. Nonetheless, the (sinusoidal) envelope observed in the pressure and volume traces in Figure 6 and in the data of Figure 7 are clearly the result of interferences that are sensitive to the phase shifts. The mean pump flow per beat does not really show this pattern because it is the primary controlled variable by the algorithm and is therefore approximately constant.
Another phenomenon that is known in wave transmission theory is the combination of waves of different frequencies. With harmonics, this results in a recurring pattern that is called “beating” in analogy with the sound formed when two such sound waves interfere. The “beat” frequency is the difference between the two original frequencies, and here it will be referred to as the “pattern frequency” to avoid confusion. In the case of a heart rate of 100 bpm (1.667 Hz) and a pump rate of 50 bpm (0.833 Hz), the pattern frequency would be 0.833 Hz, or yield a pattern period of 1.2 seconds. This is the duration of two heart beats or one pump “beat,” which agrees with the data displayed in panels B and C of Figure 4. For the combination of a heart rate of 80 bpm (1.333 Hz) and a pump rate of 60 bpm (1 Hz), a pattern period of 3 seconds occurs, which again relates to three pump beats at 60 bpm and four heart beats at 80 bpm. The 3-beat pattern can be observed in the pump flow of panel D (Figure 4), while the four-beat pattern is most apparent in the LV pressure of the same panel.
If the pattern period is not an exact multiple of the individual beat periods, then a gradual shift will occur in the pattern. The length of the pattern will remain constant, but the shape of the waveforms will change cyclically due to altered (constructive/destructive) interference. This is in fact the phenomenon observed in the HR 100 – PR 98 combination, where the pattern lasts only one beat and its variations relate to a gradual phase shift. A more apparent form of pattern changes is seen when comparing panels B and C of Figure 4, where the data were acquired during the same experiment with the same settings, but just at a different time. Apparently the rates were not exactly 100 and 50 bpm, hence the shift. The closer the heart and pump frequencies are to a multiple of the pattern frequency, the slower the shift in pattern shape will occur. Because the interaction can result in large variations in pressures and volumes, a fixed pump rate can lead to high load conditions. This in turn would imply that an automated trigger mechanism or feedback control algorithm would be necessary to avoid extreme situations.
The afterload components of the in vitro model are a simplification of the afterload as seen by the heart and the pump. Nonetheless, such windkessel models have proven their ability to simulate accurate physiological arterial pressure and flow waveforms,14–16 even though they do not allow one to study systemic perfusion on a detailed level.
The lack of baroreceptor feedback in our afterload model results in an overestimation of MAP in continuous mode (Figure 4). It can be expected in vivo that resistance decreases to normalize the MAP, which would result in an increase in flow. However, a constant total flow despite cardiac assist was previously observed in vivo in healthy animals.11,17 Due to the neglect of autonomous feedback mechanisms, the observed pressure and flow variations in pulsatile mode may appear more extreme than they will be observed in vivo. However, this means that in an in vivo setting a higher load is laid upon the nervous system and active control processes, such as the smooth muscle tone. In any case, it can be expected that stable, physiological pressure and flow waveforms are more appropriate for ventricular assist and that the here-observed unphysiological hydrodynamic variations will also be experienced as unphysiological in vivo.
The heart simulator used in this study has no elastance-based control, and consequently the sensitivity of the heart to preload (Frank-Starling mechanism) and afterload is ignored. Bypassing the original drive system with pressure feedback, however, results in a constant energy at end-diastole (buffer reservoir with fixed volume at given pressure), making the ejection afterload dependent: a lower afterload results in a higher flow (and thus stroke volume) and a lower pressure buildup. Therefore, the constant pneumatic energy that is delivered will result in a constant stroke work over a certain range of afterloads, which is similar to the in vivo findings of Glower et al.18 Afterload variations in the in vitro model will thus result in pressure volume loops as known in literature,9,19,20 similar to Figure 6.
Experiments performed with the Deltastream diagonal pump in continuous mode demonstrated that an inadequate rotational speed leads to regurgitation, the most extreme condition being pump failure. In our experiments, the negative flow effect was, however, compensated by an increase in stroke volume and cardiac output, thereby normalizing total flow. We further found that the total arterial flow is independent of the rotational speed if pump flow is kept below preimplant cardiac output.
When the same rotary pump is used in synchronous pulsatile mode, an accurate triggering with the minimal pump speed at end-systole leads to maximization of stroke volume and minimization of ventricular pressure, thus optimally unloading the ventricle. In asynchronous pulsatile mode, a complex heart-device interaction results in a repeating pattern with large pressure and flow variations. The pattern is highly unphysiological and may change slowly as a result of phase shifts. The associated risks indicate an important requirement for additional monitoring or pump control.
The authors thank Andreas Henseler and Holger Peters from Medos Medizintechnik GmbH for providing the pump and drive console. They are also grateful to Rana Shihab for her assistance with the experiment. This research was funded by a specialization grant of the Institute for the Promotion of Science and Technology in Flanders (IWT-993171, S. Vandenberghe) and by a ‘Krediet aan Navorsers’ grant of the Fund for Scientific Research in Flanders (FWO-Vlaanderen, P. Segers).
1. Christiansen S, Gobel C, Buhre W, et al: Successful use of a miniaturized bypass system with the deltastream extracorporeal rotary blood pump. J Thorac Cardiovasc Surg
125: 43–44, 2003.
2. Schmid M, Aranda-Carrero M, Christiansen S, et al: [First experiences with a miniaturized heart-lung-machine]. Biomed Tech (Berl)
47: 720–721, 2002.
3. Arvand A, Hahn N, Hormes M, et al: Comparison of hydraulic and hemolytic properties of different impeller designs of an implantable rotary blood pump by computational fluid dynamics. Artif Organs
28: 892–898, 2004.
4. Gobel C, Arvand A, Eilers R, et al: Development of the Medos/Hia Deltastream extracorporeal rotary blood pump. Artif Organs
25: 358–365, 2001.
5. Gobel C, Arvand A, Rau G, et al: A new rotary blood pump for versatile extracorporeal circulation: The Deltastream. Perfusion
17: 373–382, 2002.
6. Undar A, Lodge AJ, Daggett CW, et al: The type of aortic cannula and membrane oxygenator affect the pulsatile waveform morphology produced by a neonate-infant cardiopulmonary bypass system in vivo. Artif Organs
22: 681–686, 1998.
7. Verdonck P, Kleven A, Verhoeven R, et al: Computer-controlled in vitro model of the human left heart. Med Biol Eng Comput
30: 656–659, 1992.
8. Vandenberghe S, Segers P, Meyns B, Verdonck P: Hydrodynamic characterisation of ventricular assist devices. Int J Artif Organs
24: 470–477, 2001.
9. Suga H: Cardiac mechanics and energetics–from emax to pva. Front Med Biol Eng
2: 3–22, 1990.
10. Kono A, Maughan WL, Sunagawa K, et al: The use of left ventricular end-ejection pressure and peak pressure in the estimation of the end-systolic pressure-volume relationship. Circulation
70: 1057–1065, 1984.
11. Mihaylov D, Verkerke GJ, Blanksma PK, et al: Evaluation of the optimal driving mode during left ventricular assist with pulsatile catheter pump in calves. Artif Organs
23: 1117–1122, 1999.
12. Shiiya N, Zelinsky R, Deleuze PH, Loisance DY: Changes in hemodynamics and coronary blood flow during left ventricular assistance with the hemopump. Ann Thorac Surg
53: 1074–1079, 1992.
13. Pantalos GM, Koenig SC, Gillars KJ, et al: Characterization of an adult mock circulation for testing cardiac support devices. ASAIO J
50: 37–46, 2004.
14. Ferrari G, De Lazzari C, Kozarski M, et al: A hybrid mock circulatory system: Testing a prototype under physiologic and pathological conditions. ASAIO J
48: 487–494, 2002.
15. Sharp MK, Dharmalingham RK: Development of a hydraulic model of the human systemic circulation. ASAIO J
45: 535–540, 1999.
16. Westerhof N, Elzinga G, Sipkema P: An artificial arterial system for pumping hearts. J Appl Physiol
31: 776–781, 1971.
17. Vandenberghe S, Nishida T, Segers P, et al: The impact of pump speed and inlet cannulation site on left ventricular unloading with a rotary blood pump. Artif Organs
28: 660–667, 2004.
18. Glower DD, Spratt JA, Snow ND, et al: Linearity of the Frank-Starling relationship in the intact heart: The concept of preload recruitable stroke work. Circulation
71: 994–1009, 1985.
19. Kass DA, Maughan WL: From ‘emax’ to pressure-volume relations: A broader view. Circulation
77: 1203–1212, 1988.
20. Strackee J, Westerhof N (eds): The Physics of Heart and Circulation. Bristol, IOP Publishing Ltd., 1993.