Joint project LeRntVAD

  Copyright: © IRT

Interpretable generative machine learning for intelligent control of Ventricular Assist Devices

 

Motivation

The increasing aging population has led to a growing number of people suffering from end-stage heart failure. Due to a shortage of donor organs, the implantation of Ventricular Assist Devices (VADs) is already replacing heart transplantation as the gold standard therapy for end-stage heart failure. These systems provide mechanical support to the heart through parallel blood flow. A (physiological) control based on real-time data directly from the heart could potentially improve the quality of life for affected patients. However, current physiological control parameters require additional invasive measurements, making a near-term clinical implementation unlikely.

 

Project Goals and Methods

In LeRntVAD, a synthetic simulation model of the cardiovascular system is developed based on existing experimental data. The scientific challenge lies in the intuitive access to generative models of machine learning, which is necessary to determine the type of synthetic data to be generated. The scientific goal is to design a controller based on artificial intelligence (AI) for left-sided heart support based on the data generated by the generative model, thereby expanding the existing controllers to eliminate the need for further invasive measurements.

The IRT has been engaged in the design of intelligent, adaptive, and model-based controllers for Ventricular Assist Devices for many years. Data-driven controllers represent the next logical and desirable development stage. Data-driven methods have so far been applied to identification and modeling. The integration of such a methodology into the closed control loop is of utmost interest. The scientific objective of the project is to successfully test the developed AI-controller as part of a closed control loop.

 

Project partner
Consortium Coordinator: RWTH Aachen University, IRT - Institut für Regelungstechnik Bielefeld University, Machine Learning Group University Hospital Aachen, Department of Anesthesiology