MEG Source Localization via Deep Learning

TL;DR

This paper introduces a deep learning approach for MEG source localization that outperforms traditional methods in accuracy and speed, enabling real-time brain activity mapping with robustness to model errors.

Contribution

The paper develops novel deep learning architectures tailored for MEG data that improve localization accuracy and significantly reduce computation time compared to existing algorithms.

Findings

Outperforms RAP-MUSIC in simulated scenarios

Achieves localization in under 1 millisecond

Robust to forward model errors

Abstract

We present a deep learning solution to the problem of localization of magnetoencephalography (MEG) brain signals. The proposed deep model architectures are tuned for single and multiple time point MEG data, and can estimate varying numbers of dipole sources. Results from simulated MEG data on the cortical surface of a real human subject demonstrated improvements against the popular RAP-MUSIC localization algorithm in specific scenarios with varying SNR levels, inter-source correlation values, and number of sources. Importantly, the deep learning models had robust performance to forward model errors and a significant reduction in computation time, to a fraction of 1 ms, paving the way to real-time MEG source localization.