Real-Time Integration of Stochastic LVAD Models

Integrate stochastic cardiovascular models into real-time left ventricular assist device (LVAD) control systems despite computational constraints and the lack of reliable implantable sensors for continuous monitoring.

Background

The paper explains that stochastic modeling approaches are intended to represent uncertainty arising from inter-patient variability, intra-patient physiological fluctuations, measurement noise, and unmodeled disturbances in LVAD-supported circulation. Examples discussed include Monte Carlo sampling, Gaussian Process models, and generalized polynomial chaos expansions for uncertainty quantification, physiological estimation, and self-tuning control.

Although these methods can improve robustness under changing physiological conditions, the paper identifies their integration into real-time LVAD control as unresolved. The principal barriers are the computational cost of stochastic methods and the absence of reliable implantable sensors capable of continuously measuring the physiological information required by such controllers.