Neural networks can feasibly reconstruct cardiac currents from simulated magnetocardiograms using a forward model.
Answers are grounded in this paper and cite their sources.
Neural network feasibility in simulation supports further tool development; leaves open clinical translation and validation in patients.
The biomagnetic reconstruction of cardiac currents with a trained neural network was performed. The input-output pairs for training the neural network consisted of the magnetic field profiles above the torso and their timings in the heartbeat. A back-propagating algorithm with momentum and bias was utilized for training. The cardiac magnetic field profiles and magnetocardiograms (MCGs) for training the neural network were computed using a realistic, inhomogeneous, finite element model of the heart and torso. The electrical activity of the heart was modelled with distributed dipole sources. Reconstruction of the sources for unknown magnetic field profiles were performed with the trained neural network. The reconstructed cardiac dipole sources were similar to the true dipoles. This shows the feasibility of the biomagnetic reconstruction of cardiac currents with neural networks and a forward model of the cardiac magnetic fields.
No takes yet. Share an insight, caveat, or question.
Ceon Ramon (1997) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: