Fully Autonomous Preload‐Sensitive Control of Implantable Rotary Blood Pumps

Abstract

A pulsatility-based control algorithm with a self-adapting pulsatility reference value is proposed for an implantable rotary blood pump and is to be tested in computer simulations. The only input signal is the pressure difference across the pump, which is deduced from measurements of the pump's magnetic bearing. A pulsatility index (PI) is calculated as the mean absolute deviation from the mean pressure difference. As a second characteristic, the gradient of the PI with respect to the pump speed is derived. This pulsatility gradient (GPI) is used as the controlled variable to adjust the operating point of the pump when physiological variables such as the systemic arterial pressure, left ventricular contractility, or heart rate change. Depending on the selected mode of operation, the controller is either a linear controller or an extremum-seeking controller. A supervisory mechanism monitors the state of the system and projects the system into the region of convergence when necessary. The controller of the GPI continuously adjusts the reference value for PI. An underlying robust linear controller regulates the PI to the reference value in order to take into account changes in pulmonary venous return. As a means of reacting to sudden changes in the venous return, a suction detection mechanism was included. The control system is robustly stable within a wide range of physiological variables. All the clinician needs to do is to select between the two operating modes. No other adjustments are required. The algorithm showed promising results which encourage further testing in vitro and in vivo.

Automatic control of dual left ventricular assist devices as a biventricular assist device

“…Performance metrics included observing the settling time, the number of suction events that occurred and time to resolve suction events and restore normal operation. The response time of the inner loop was fast, with the PI restored to its initial value within 25 seconds, although this was reduced to 15 with the implementation of dedicated suction avoidance strategies [83]. The outer loop was much slower, with the system taking 500 seconds to switch from FA to partial assist mode.…”

“…Notable omissions included work by Arndt et al [82], [83], Vollkron and colleagues [84], Choi et al [80], [95] and Gwak et al [86], [87]. These controllers were omitted because their complexity, in conjunction with insufficient detail of controller parameters in their published descriptions, made it difficult to implement them in this experimental setup within the time frame.…”

“…As Waters et al [92] and Arndt et al [83] lumped right heart and pulmonary circulatory elements together as a single pressure source, LV preload was changed simply by adjusting the value of this pressure source. Gwak andcolleagues (2007, 2011) did not use a traditional lumped parameter model for evaluation, so these investigators empirically varied preload by changing the horizontal and vertical offset of their cost function for control optimisation [86], [87].…”

“…locating where the second derivative is zero ( Figure 4.11, bottom) [82], [83], [87]. Gwak and colleagues [87] compared this approach using ΔP pulsatility and minimum pump flow approaches, using a simplified NM for evaluation (previously described in [86]).…”

“…A solution proposed both by Arndt et al (2008Arndt et al ( ,2010 and Gwak et al (2011) to determine the optimal slope for a ΔP pulsatility vs. pump speed curve (Figure 4.11, top) was to use ESC to operate at the minimum of the first derivative of this curve i.e. locating where the second derivative is zero ( Figure 4.11, bottom) [82], [83], [87].…”