Key result
Different sequence autoencoder architectures focus on different aspects of cardiac transmembrane potential, with stochastic models better learning underlying factors like scar regions compared to deterministic language models.
Why the study?
Prior and conditional distributions learned by deep generative models for inverse electrophysiological imaging are directly affected by neural network architecture, but the effect of architecture on representation learning and generalization remains to be investigated.
The choice of neural network architecture in deep generative models significantly affects the reconstruction and representation of cardiac transmembrane potentials from ECG data.
Architecture choice may influence ECG-based scar mapping; leaves open optimal generative models for clinical translation.
It has been shown recently that inverse electrophysiological imaging can be improved by using a deep generative model learned in an unsupervised way so that cardiac transmembrane potential and underlying generative models could be simultaneously inferred from the ECG. The prior and conditional distributions learned in such a way are, however, directly affected by the architecture of neural network used in unsupervised learning. In this paper, we investigate the effect of architecture in learning representation and generalizing to new test cases. By comparing reconstruction of three types of sequence autoencoder, we show that different sequence autoencoders might be focusing on different aspects of TMP and might perform differently according to the metric used to measure reconstruction. We also analyze the latent space in different architectures and discuss important questions raised by these observations.
No takes yet. Share an insight, caveat, or question.
Ghimire et al. (2018) studied Cardiac Transmembrane Potential (simulation) (n=600). Sequence autoencoder architectures (Language, svs, sss) vs. Comparison among architectures was evaluated on Reconstruction accuracy (MSE, correlation, dice coefficient, activation time correlation). Different sequence autoencoder architectures focus on different aspects of cardiac transmembrane potential, with stochastic models better learning underlying factors like scar regions compared to deterministic language models.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: