Cardiac biomagnetic source estimation with a heart-torso model and a trained neural network

Area of Science:

• Biophysics
• Computational Neuroscience

Background:

• Cardiac magnetic fields reflect the heart's electrical activity.
• Accurate estimation of cardiac sources is crucial for diagnosing heart conditions.

Purpose of the Study:

• To develop and validate a neural network model for estimating cardiac dipole intensities from magnetic field profiles.
• To assess the feasibility of using machine learning for non-invasive cardiac source analysis.

Main Methods:

• A neural network was trained using measured and simulated torso magnetic field profiles and magnetocardiograms.
• A backpropagation algorithm with bias and momentum was employed for network training.
• The network learned the relationship between magnetic field profiles and underlying dipole intensities.

Main Results:

• The trained neural network accurately estimated cardiac dipole intensities from unknown magnetic field profiles.
• Estimated dipole intensities closely matched true dipole intensities in validation tests.
• The model demonstrated the feasibility of using neural networks for cardiac magnetic field analysis.

Conclusions:

• Neural network-based estimation of cardiac dipole intensities is feasible under simplified forward models.
• This approach offers a potential tool for non-invasive analysis of cardiac electrical activity.
• Wider clinical application may require addressing inter-subject variability in cardiac activation patterns.