Abstract

Heart failure with preserved ejection fraction (HFpEF) represents a growing subset of cardiovascular disease, characterized by impaired ventricular relaxation despite normal systolic function. Conventional left ventricular assist devices (LVADs), designed primarily for systolic dysfunction (HFrEF), fail to provide physiological assistance for diastolic filling deficiencies typical of HFpEF. This study presents the modeling, simulation, and control of a novel LVAD architecture designed to aid the left atrium-to-ventricle blood transfer during diastolic stroke. A comprehensive sixth-order CVS-LVAD model is developed based on lumped parameter representation, capturing the nonlinear hemodynamic behavior under varying preload and afterload conditions. The model’s validity is confirmed by reproducing physiological pressure–volume relationships and cardiac output trends across resting and active states. To regulate the LVAD, a two-tier intelligent Model Predictive Controller (MPC) is designed. The upper tier employs a machine learning (ML) module, trained using real patient data, to estimate systemic vascular resistance (SVR), cardiac output (CO), and a dynamic reference based on patient-specific parameters such as age, height, weight, and heart rate. Among five ML algorithms tested (SVR, ANN, GAM, LSBoost, and RFR), Support Vector Regression (SVR) achieved the highest accuracy (R² > 0.90) and computational efficiency, thus selected for integration with MPC. The MPC ensures real-time tracking of cardiac flow trajectories by adjusting pump speed during diastolic strokes while maintaining hemodynamic safety constraints. Simulation results under varying activity conditions demonstrate that the proposed ML-MPC controller effectively synchronizes pump assistance with cardiac cycles, maintains mean arterial pressure (MAP) and cardiac output within physiological limits (6-11.2 L/min), and prevents suction or backflow. The framework establishes a foundation for adaptive, patient-specific LVAD control tailored for HFpEF, marking a significant step toward intelligent circulatory support systems.

About the Speaker

Mr. Ashhad Ali Shahid is an Engineering Student with interests in control of nonlinear systems, robotics, and biomedical technologies. He is currently pursuing his M.Sc. in Systems and Control Engineering in the Control and Instrumentation Engineering Department at King Fahd University of Petroleum and Minerals (KFUPM), Saudi Arabia. He obtained his B.Sc. in Mechanical Engineering from National University of Sciences and Technology, Islamabad, Pakistan. Mr. Ashhad Ali has worked as a design engineer for a short period at National Aerospace Science and Technology Park, Islamabad. His research interests include modelling of mechanical systems, nonlinear systems, neural networks, control in biomedical applications, and intelligent control algorithms. Mr. Ashhad’s contributions include his work in IoT-based digital twins for vibrating beams, structural design of aircraft, modelling of the cardiovascular system, and development of sensing and control algorithms for LVAD pumps.