Data-driven discovery of sparse dynamical model of cardiovascular system for model predictive control

Area of Science:

• Biomedical Engineering
• Cardiovascular Research
• Neuroscience

Background:

• Cardiovascular diseases are a leading global cause of death.
• Vagal nerve stimulation (VNS) shows promise for treating cardiovascular conditions.
• Current open-loop VNS has limitations due to patient variability, necessitating closed-loop strategies.

Purpose of the Study:

• To develop a data-driven dynamical model for cardiovascular responses to VNS.
• To design a closed-loop controller for VNS using the developed model.
• To assess the model's adequacy for optimizing VNS in cardiovascular disease treatment.

Main Methods:

• Utilized sparse identification of nonlinear dynamics (SINDy) to build a dynamical model.
• Simulated patient data using a mechanistic model as a proxy for real measurements.
• Designed a model predictive control (MPC) strategy for closed-loop VNS implementation.

Main Results:

• A data-driven dynamical model for mean arterial pressure and heart rate was successfully built.
• The discovered model demonstrated interpretability and consistency with experimental data.
• The closed-loop MPC controller using the data-driven model achieved effective set-point tracking.

Conclusions:

• The data-driven model adequately captures the nonlinear dynamics of the cardiovascular system for VNS control.
• This approach offers a pathway for personalized and optimized VNS therapy.
• The findings support the use of data-driven models in designing advanced neurostimulation controllers.