(284g) Model Free Adaptive Control of the Failing Heart Managed By Mechanical Supporting Devices

Heart failure (HF) is a syndrome that the heart cannot provide sufficient blood out of the left ventricle as it should. It is estimated that about 6.5 million patients in the U.S. undergo HF [1]. Although heart transplantation is recognized as the best therapeutic option for advanced HF, only a few patients can receive heart transplants due to the limited number of available organ donors. Left ventricular assist device (LVAD) is one of the well-known clinical therapies for advanced HF patients. As a mechanical circulatory device, an LVAD is implanted in the patient to support the blood circulation by pumping blood from the failing left ventricle to the aorta [2].

Currently, the pump speed of the LVAD is set in a constant mode by clinicians and cannot be freely adjusted. However, hemodynamic of the cardiovascular system varies over time (e.g., changes in heart condition and levels of physical activities). Thus, the constant speed-based operation of an LVAD can lead to detrimental events such as insufficient perfusion or ventricular suction. To promote the broader application of an LVAD as a long-term treatment option, a physiological control system is needed to automatically adjust the pump speed in response to hemodynamic variations.

Many physiological control algorithms have been designed in the literature to adjust the pump speed of an LVAD according to the physiologic needs of HF patients. However, most of the previously developed controllers have fixed gains to execute physiological control systems, which can provide the desirable control performance with a limited range of hemodynamic of a patient. Recently, adaptive control algorithms have been proposed, which can adjust control parameters in response to feedback from the controlled system. This includes artificial neural network (ANN) control [3], fuzzy logic control (FLC) [4], gain-scheduling control [5], and self-tuning control [6]. Most of the existing works require an accurate model of the heart for controller tuning, but it is difficult to model the heart using first principles and to identify the model parameters with measured data. Since data contains valuable information, a model free adaptive control (MFAC) [7] was recently used to control LVADs [8]. However, the compact linearization was used for MFAC to adaptively adjust the pump speed of an LVAD. In this case, the control performance cannot be guaranteed since the hemodynamics of the heart cannot be identified properly with linearized models [7], [8]. Thus, it is necessary to account for uncertainty in MFAC control system to obtain the desired control performance. Another challenge to design a control algorithm for an LVAD is the appropriate selection of the feedback variable. Previous studies on the physiological control systems often rely on an implantable sensor to obtain direct measurements (e.g., left ventricular pressure, ventricular volume, and blood flow) for control. However, the implantation of sensors may cause complications such as thrombus formation and device failure, and it is not suitable for long-term applications.

To design a physiological control system, a lumped mathematical model of the human cardiovascular system managed by an LVAD is commonly used to predict hemodynamics, due to the limited accessibility to clinical data. These models generally include the systemic circulation, pulmonary circulation, both sides of the heart, and the LVAD. While useful, the application of these models in clinics is limited due to the lack of information about the hemodynamic caused by baroreflex regulation. Baroreflex maintains blood pressure, causes heart rate to change, and is one of the key factors of human homeostatic mechanisms. Therefore, it is necessary to integrate it with the failing heart managed by an LVAD to advance our understanding of the cardiovascular system.

In this work, a new adaptive control strategy is proposed to automatically adjust the pump speed of the LVAD using MFAC. To test our algorithm, hemodynamic of the heart is simulated with the cardiovascular system managed by an LVAD, while considering the baroreflex regulation. Using the synthetic data generated from the models, an empirical model is developed to invasively estimate the controlled variable in real-time based on the relationship between the pulmonary venous deceleration time and the left ventricular end-diastolic pressure (LVDEP) [9]. To adjust the pump speed, an MFAC-based controller is designed to adjust the pump speed of the LVAD. The efficiency of our algorithm is tested with different conditions of patients through computer simulations, thus laying the foundation for clinical application of LVADs.

References

[1] E. J. Benjamin et al., “Heart disease and stroke statistics—2018 update: a report from the American Heart Association,” Circulation, vol. 137, no. 12, pp. e67–e492, Mar. 2018.

[2] A.-H. H. AlOmari et al., “Developments in control systems for rotary left ventricular assist devices for heart failure patients: a review,” Physiol. Meas., vol. 34, no. 1, pp. R1–R27, Dec. 2012.

[3] B. C. Ng et al., “Application of multiobjective neural predictive control to biventricular assistance using dual rotary blood pumps,” Biomed. Signal Process. Control, vol. 39, pp. 81–93, Jan. 2018.

[4] T. Leao, B. Utiyama, J. Fonseca, E. Bock, and A. Andrade, “In vitro evaluation of multi-objective physiological control of the centrifugal blood pump,” Artif. Organs, vol. 44, no. 8, pp. 785–796, Aug. 2020.

[5] Y. Wang, S. C. Koenig, M. S. Slaughter, and G. A. Giridharan, “Rotary blood pump control strategy for preventing left ventricular suction,” ASAIO J., vol. 61, no. 1, pp. 21–30, Jan.-Feb. 2015.

[6] J. Son, D. Du, and Y. Du, “Modelling and control of a failing heart managed by a left ventricular assist device,” Biocybern. Biomed. Eng., vol. 40, no. 1, pp. 559–573, Jan. 2020.

[7] Z. Hou and S. Jin, Model Free Adaptive Control: Theory and Applications. Taylor & Francis, 2013.

[8] M. Fetanat, M. Stevens, C. Hayward, and N. H. Lovell, “A Physiological Control System for an Implantable Heart Pump That Accommodates for Interpatient and Intrapatient Variations,” IEEE Trans. Biomed. Eng., vol. 67, no. 4, pp. 1167–1175, Apr. 2020.

[9] A. Olariu, E. Wellnhofer, M. Gräfe, and E. Fleck, “Non-invasive estimation of left ventricular end-diastolic pressure by pulmonary venous flow deceleration time,” Eur. J. Echocardiography, vol. 4, no. 3, pp. 162–168, Sep. 2003.