Better prediction of heart failure with data and computer models

Ventricular tachycardia (VT or V-tach) is a type of abnormal heart rhythm. In the hospital there is a lot of data, including echo, MRI, ECG and laboratory values, from patients with this disease that is currently not used. In a collaboration between Eindhoven University of Technology (TU/e), Philips Research and Catharina Hospital, the COMBAT-VT project aims to develop a computational model that uses the patient’s data to predict and support the risk of cardiac arrhythmia the decision for the best treatment. According to a TU/e ​​press release, this would significantly improve the quality of life while reducing healthcare costs.

Patient data is used for treatment, but there is also physiological and physical knowledge in the form of computer models. The project will look at Physics Informed Neural Networks (PINNs) trying to train on the data while respecting the laws of physics. This makes the training more efficient.

Electromechanical modelling

Electrophysiological VT simulations rely on geometric information of the infarcted area, and the remodeling of tissue properties in the current electrophysiological modeling pipeline is limited to this area. However, in mechanical simulations, functional abnormalities are also detected outside of this infarcted area, where deformations of the nearby distant area are affected by the dysfunctional infarct.

In this project, such mechanical anomalies are linked to electrophysiological remodeling. Here, abnormal stress distributions across the heart are hypothesized to affect the redistribution of remodeled electrophysiological properties over time. The implications of this hypothesis for long-term computational VT risk predictions will be further investigated to see if it could explain the progression of VT risk over time as observed in the clinic.

Reduce the computational effort

The new project also aims to use a new computational model for analyzing VT data called Isogeometric Analysis (IGA). IGA allows the construction of smooth discretized geometries. It is expected to reduce the computational effort compared to currently used methods while providing an accurate representation of the evolution and behavior of VTs.

Leave a Comment