Share this article on:

A Safe Automatic Driving Method for a Continuous Flow Ventricular Assist Device Based on Motor Current Pulsatility:In Vitro Evaluation

Endo, George; Araki, Kenji; Oshikawa, Mitsuo; Kojima, Kazushi; Nakamura, Kunihide; Matsuzaki, Yasunori; Onitsuka, Toshio

Original Articles

We previously reported that detection of two specific points (the t-point and the s-point) in the relationship between pump speed and Motor Current Amplitude index (ICA) indicates the safe driving range for a continuous-flow ventricular assist device (CFVAD). During the first stage of the present experiment, the characteristic curves relating pump speed and ICA were determined by varying preload (left atrial pressure: −6 to 30 mm Hg), afterload (total circuit resistance: 890 to 3,180 dyne · sec · cm−5), and contractility of the left ventricle (total circuit flow: 0.5 to 2.1 L/min). These data showed that an ICA value of 0.18 was always located between the t- and s-points. During the second stage of the experiment, we developed an automatic driving program to control pump speed by maintaining ICA at 0.18. This program was able to drive the CFVAD, without exhibiting regurgitant flow or sucking, under various driving conditions in the mock circulation. Pump speed stabilized within 1 minute after varying the drive conditions. This sensorless method of driving the CFVAD by using a target ICA proved feasible and effective for safe automatic control, within our mock circulation.

From the Miyazaki Medical College, Miyazaki, Japan.

Submitted for consideration June 2000; accepted for publication in revised form June 2001.

Reprint requests: Dr. George Endo, 2nd Department of Surgery, Miyazaki Medical College, 5200 Kihara, Kiyotake Miyazaki 889-1692 Japan.

A ventricular assist device (VAD) provides a method for rescuing severe heart failure cases, such as those with dilated cardiomyopathy and postoperative low output syndrome. Recently, much attention has been paid to the use of a VAD as a bridge to transplantation or for permanent circulatory support, following social pressure in response to the shortage of donor organs. Continuous-flow pumps are preferable as the driving mechanism, owing to their endurance, reduced size, and cost performance. On the other hand, a continuous flow pump can never be allowed to stop once it has been implanted, and requires some method for controlling the pump speed and, hence, pump flow.

We previously reported that two specific points detected on a graph of the relationship between pump speed and ICA, called the t-point and s-point (Figure 1), were indicative of the safe range for driving the CFVAD. 1,2 When a target value of ICA (target ICA), which is located between these two points, can be identified under conditions of varying preload, afterload, and contractility, the CFVAD can be safely driven. In this study, we investigated the feasibility of using the target ICA for safe operation and developed a prototype automatic driving program with the aid of a closed-loop mock circulation.

Back to Top | Article Outline

Materials and Methods

Mock Circulation

Our closed-loop circuit consists of a sac-type pulsatile pump simulating the natural left ventricle (maximum stroke volume, 40 ml; drive pressure, 300 mm Hg; vacuum pressure, −100 mm Hg; pulse rate, 60 bpm), three reservoirs, representing the arterial/venous vascular bed and the left atrium, and a mixed-flow pump in the role of the CFVAD. The preload can be changed by altering the filling volume of the circuit, whereas the afterload can be changed by controlling a throttle on the proximal side of the arterial reservoir. Contractility can be changed by controlling a throttle in the pneumatic drive line (Figure 2). In the baseline circulating state without the CFVAD, we set the mean left atrial pressure (LAP) to 10 mm Hg, the mean aortic pressure (AoP) to 60 ± 5 mm Hg, and the total circuit flow (TCF) to 2.0 ± 0.4 L/min.

Back to Top | Article Outline


Our mixed-flow pump is characterized by a fully open impeller with six curved, twisted vanes (a three-dimensional structure), the maximum diameter of which is 32 mm. The height of the pump case is 30 mm, and its maximum diameter is 42 mm, with a priming volume of 15 ml. A magnetic coupling mechanism eliminates the need for a sealed drive shaft. The impeller is driven with a brushless DC motor.

Back to Top | Article Outline

Sampling and Recording of Data

AoP, left ventricular pressure (LVP), LAP, CFVAD pump speed, CFVAD motor current, TCF, and CFVAD flow rate were measured. ICA was calculated from the motor current waveform and the simultaneous value of mean current. The formula for ICA is as follows:MATH where I max, I min, and I mean are the maximum, minimum, and mean values of the motor current waveform, respectively.

All values were digitized for input into a personal computer (VT516SV; EPSON, Tokyo, Japan) and recorded using a data acquisition program (Labview 5.0, National Instruments, Austin, TX). Pressures were input by means of a polygraph (EP1102; NEC, Tokyo, Japan). Motor current and pump speed were input directly.

Back to Top | Article Outline


Experiment 1. Characteristic Curves of Pump Speed Versus ICA under Various Cardiac Conditions.

The pump speed of the CFVAD was varied from 0 to 5,000 rpm or until sucking occurred, the data being sampled and recorded at 10 to 15 steps within this range. The ICA value was then plotted to graphically depict the relationship. The cardiac conditions were changed as follows: (1) preload - mean left atrial pressure (mLAP) from −6 to 30 mm Hg (15 steps); (2) afterload - total circuit resistance (TCR) from 890 to 3,180 dyne · sec/cm5 (10 steps); contractility - total circuit flow (TCF) from 0.5 to 2.1 L/min (12 steps).

Back to Top | Article Outline

Experiment 2. Automatic Driving Program.

We expected to obtain a feasible control method for the CFVAD, using the ICA, by one or more of the following methods: (1) keeping the ICA at a fixed value, (2) detecting the slope of the ICA curve, (3) referring to the minimum ICA, (4) referring to the maximum ICA. We believed that the most convenient method would use (2) and (3) but selected the fixed-ICA mode in this study due to fastest hardware response. The details of this algorithm are shown in Figure 3. In this algorithm the feedback loop ran every 3 seconds, changing the pump speed by 50 rpm, executed every 15 seconds by hardware feedback-delay. After executing the program, sampling data were recorded every 15 seconds, whereas the afterload or contractility was rapidly changed at six 2 minute intervals (TCR from 890 to 2,800 dyne · sec/cm5, base TCF as 0.15 to 2.0 L/min, respectively)

In this study, when sucking occurred, the motor current waveform pattern showed a downward spike. Once this was established, the prototype program for detecting sucking, which relies on the motor current waveform, was installed and run during both experiments.

Back to Top | Article Outline


Relationship between Pump Speed and ICA


As preload was increased, pump speed at the t-point remained almost constant (2,125.9 ± 206.3 rpm) and that at the s-point increased from 2,109.4 to 4,087.6 rpm. Similarly, pump flow remained almost constant at the t-point (4.42 ± 0.29 L/min) but increased at the s-point from 0.34 to 7.07 L/min. The pump flow showed a tendency to be preload dependent, but when the mLAP fell below −5 mm Hg, the pump flow at both points dropped rapidly (from 3.42 to 0.25 L/min at the t-point, and from 3.78 to 0.34 L/min at the s-point). The ICA value at the t-point (ICA-t) gradually increased from 0.20 to 0.36, and that at the s-point (ICA-s) gradually decreased from 0.16 to 0.06 but remained almost constant. These changes provided a wide safety range in the pump speed versus ICA relationship (Figure 4).

Back to Top | Article Outline


As afterload was increased, pump speed at the t-point remained almost constant, with a slight tendency to increase (2,689.3 ± 289.6 rpm), and that at the s-point increased from 2,902.6 to 4,675.7 rpm. Pump flow at the t-point gradually decreased from 2.77 to 4.97 L/min, and that at the s-point remained almost constant (5.01 ± 0.39 L/min).

The ICA-t increased from 0.21 to 0.63, while the ICA-s gradually decreased from 0.15 to 0.05, remaining almost constant. These changes shifted the pump speed versus ICA curve to the right (Figure 5).

Back to Top | Article Outline


As afterload was increased, pump speed at the t-point increased from 1,206 to 3,112.9 rpm, while that at the s-point remained almost constant (3,130.3 ± 286.8 rpm). Similarly, pump flow at the t-point increased from 1.14 to 4.87 L/min, while that at the s-point remained almost constant (4.85 ± 0.38 L/min).

The ICA-t gradually increased from 0.22 to 0.52, and the ICA-s gradually decreased from 0.17 to 0.07, remaining nearly constant. These changes yielded a narrow safety range in the pump speed versus ICA curve (Figure 6). In all experiments, the plotted values of ICA at the t-point exceeded 0.2, while those at the t-point were below 0.17.

From these results, it was clear that the CFVAD could be maintained in a total-assist state, in this mock circulation, for an ICA of 0.18 or 0.19. At an ICA of 0.2 ± 0.03, the pump flow remained almost constant when the preload changed (4.35 ± 0.07 L/min), when the afterload changed (3.95 ± 0.20 L/min), and when the contractility changed (4.16 ± 0.09 L/min), although the changes in pump speed were different (2,495 ± 28 rpm, 2,841 ± 122 rpm, and 2,544 ± 78 rpm, respectively).

Back to Top | Article Outline

Automatic Driving Program

From the results of experiment 1, we set the target ICA as 0.18 and executed the autopilot program. After a rapid change in afterload or contractility, the ICA value also changed rapidly. Pump speed responded immediately to the change in ICA, the value of which converged with the target ICA within 1 minute. Consequently, pump flow remained constant, although pump speed increased beyond the higher safety limit when the TCR gain was over 2,800 dyne · sec/cm5 or the base TCF dropped to 150 ml/min (Figure 7).

Back to Top | Article Outline


Currently, continuous flow pumps are used for intraoperative cardiopulmonary bypass and extracorporeal life-support systems, but these devices are usually manually controlled. 3–5 Because clinical application of the LVAD or RVAD commonly involves a pulsatile pump, 6–12 control strategy of the CFVAD has received little discussion.

In our previous report, the ICA was shown to be able to detect the total assist state at the t-point and ventricular collapse at the s-point, in a beating heart supported by a continuous flow pump. 1,2 The ICA was able to delineate the safety range of the LVAD speed so as to avoid both regurgitant flow and sucking. 13–16 However, in that series of experiments, the target ICA was determined empirically and the VADs were controlled manually. In the present series, we evaluated the characteristics of the ICA curve under various cardiac states and successfully deployed a prototype automatic control program.

In this mock circulation, when mLAP fell below −6 mm Hg, the pump could not maintain sufficient flow because the reservoir could not be completely filled. With the exception of this extreme condition, changes in pump flow occurred in a preload-dependent manner, as we previously reported. Because the sucking phenomenon is related to the cavity volume and is independent of ventricular contractility, the s-point was not affected by contractility but was somewhat dependent on preload and afterload. Following this tendency, the value of the ICA-s remained almost constant.

In experiment 2, the pump flow was able to remain almost constant in the presence of rapid changes in afterload or contractility by monitoring the ICA value and varying pump speed to attain the target ICA. These results suggest that this program is capable of controlling the CFVAD in conscious animals. Because this program was only designed for controlling the pump within the safety range, the animals’ state was not a consideration, i.e., the pump flow that is ideal for the pump might not match the animal’s circulatory requirements. To avoid such a mismatch, other input parameters are required, such as mixed venous saturation. 17,18 In addition to such monitoring, and although a pump flowmeter is not always required for circulatory assistance, estimation of pump flow 19 or a micro-flowmeter should be included to improve the safety of the drive system in animals.

This program requires manual input of the upper and lower limits of pump speed as well as the target ICA for safe operation. Therefore, this control method is semiautomatic, not fully automatic. Although the prototype of the sucking detection program performed to expectation, its accuracy is still imperfect. At an early date, an improved version of this program might replace the manual input of the upper pump speed limit.

However, the method for fixing the target ICA requires further consideration. In this series, we determined the target ICA from the pump speed/ICA curve, but, for animals, the pump speed should not be allowed to approach the extreme conditions of regurgitation or sucking. From the results of experiment 1, the fact that the ICA-s remained almost constant demonstrates the importance of this parameter. We are developing an automatic set-up program that uses autodetection of sucking to establish the target ICA. On the other hand, low preload or low afterload conditions lead to low ICA-t, which approaches the value of ICA-s. This condition could cause overshoot of the control output even if the target ICAs were accurately established. To avoid this problem, we are developing a program that detects the slope of the ICA curve and determines the pump speed within the safety range.

Back to Top | Article Outline


The ICA was able to describe the safety range of LVAD speed to avoid regurgitant flow and sucking. Setting an accurate target ICA, as established from the characteristics of the ICA curve under various cardiac conditions, our autopilot program was able to restrict pump speed to the safety range and maintain constant pump flow, without feedback from a flow meter. We conclude that a control method using the ICA is feasible for controlling the CFVAD and may ultimately enable fully automatic control.

Back to Top | Article Outline


1. Araki K, Oshikawa M, Onitsuka T, Nakamura K, Anai H, Yoshihara H: Detection of total assist and sucking points based on pulsatility of continuous-flow artificial heart: in vitro evaluation. ASAIO J 44: M708–M711, 1998.
2. Araki K, Oshikawa M, Anai H, et al: Detection and implication of the characteristic points for estimating the assisting condition in a continuous-flow artificial heart: In vitro study. Jpn J Artif Organs 28: 383–393, 1999.
3. Noon GP, Lafuente JA, Irwin S: Acute and temporary ventricular support with Biomedicus centrifugal pump. Ann Thorac Surg 68: 650–654, 1999.
4. Cyrtis JJ, Walls JT, Wagner-Mann CC, et al: Centrifugal pumps: Description of devices and surgical techniques. Ann Thorac Surg 68: 666–671, 1999.
5. Walter PD: Bridging from acute to chronic devices. Ann Thorac Surg 68: 666–671, 1999.
6. Morales DLS, Catanese KA, Helman DN, et al: Six-year experience of caring for forty-four patients with a left ventricular assist device at home: safe, economical, necessary. J Thorac Cardiovasc Surg 119: 251–259, 2000
7. Sun BC, Catanese KA, Spanier TB, et al: 100 long-term implantable left ventricular assist devices: The Columbia Presbyterian interim experience. Ann Thorac Surg 68: 668–694, 1999.
8. Robbins RC, Oyler PE: Bridge to transplant with the Novacor left ventricular assist system. Ann Thorac Surg 68: 695–697, 1999.
9. Copeland JG, Arabia FA, Smith GS, Sethi GK, Nolan PE, Bancht ME: Arizona experience with CardioWest total artificial heart bridge to transplantation. Ann Thorac Surg 68: 756–760, 1999.
10. Nose Y, Ohtsubo S, Tayama E: Therapeutic and physiological artificial heart: future prospects. Artif Organs 21: 592–596, 1997.
11. Wudel JH, Hlozek CC, McCarthy PM: Extracorporeal life support as a post left ventricular assist device implant supplement. ASAIO J 43: M441–M443, 1997.
12. Hall AW, Soykan O, Harken AH: Physiologic control of cardiac assist devices. Artif Organs 20: 271–275, 1996.
13. Oshikawa M, Araki K, Nakamura K, Anai H, Onitsuka T: Detection of total assist and sucking points based on pulsatility of continuous-flow artificial heart: in vivo evaluation. ASAIO J 44: M704–M707, 1998.
14. Oshikawa M, Araki K, Anai H, Nakamura K, Satoh M, Onitsuka T:In vivo evaluation of novel control method for a continuous-flow artificial heart. Jpn J Artif Organs 28: 320–325, 1999.
15. Endo G, Araki K, Oshikawa M, et al: Control strategy for biventricular assistance with mixed-flow pump:Artif Organs (in press).
16. Oshikawa M, Araki K, Endo G, Onitsuka T, Yoshihara Y, Anai H: Pulsatility-based control for rotary blood pumps evaluated in dogs, pigs and goats:Artif Organs (in press).
17. van der Hoeven MA, Maertzdorf WJ, Blanco CE: Mixed venous oxygen saturation and biochemical parameters of hypoxia during progressive hypoxemia in 10- to 14-day-old piglets. Pediatr Res 42: 878–884, 1997.
18. Nakamura M, Masuzawa T, Tatsumi E, et al: The development of a control method for a total artificial heart using mixed venous oxygen saturation. Artif Organs 23: 235–241, 1999.
19. Tsukiya T, Akamatsu T, Nishimura K, Yamada T, Nakazeki T: Use of motor current in flow rate measurement for the magnetically suspended centrifugal blood pump. Artif Organs 21: 396–401, 1997.
Copyright © 2002 by the American Society for Artificial Internal Organs