Myelin-informed forward models for M/EEG source reconstruction
Preprint
Abstract
Magnetoencephalography (MEG) and Electroencephalography (EEG) provide direct electrophysiological measures at an excellent temporal resolution, but the spatial resolution of source-reconstructed current activity is limited to several millimetres. Here we show, using simulations of MEG signals and Bayesian model comparison, that non-invasive myelin estimates from high-resolution quantitative magnetic resonance imaging (MRI) can enhance MEG/EEG source reconstruction. Our approach assumes that MEG/EEG signals primarily arise from the synchronised activity of pyramidal cells, and since most of the myelin in the cortical sheet originates from these cells, myelin density can predict the strength of cortical sources measured by MEG/EEG. Leveraging recent advances in quantitative MRI, we exploit this structure-function relationship and scale the leadfields of the forward model according to the local myelin density estimates from in vivo quantitative MRI to inform MEG/EEG source reconstruction. Using Bayesian model comparison and dipole localisation errors (DLEs), we demonstrate that adapting local forward fields to reflect increased local myelination at the site of a simulated source explains the simulated data better than models without such leadfield scaling. Our model comparison framework proves sensitive to myelin changes in simulations with exact coregistration and moderate-to-high sensor-level signal-to-noise ratios (≥10 dB) for the multiple sparse priors (MSP) and empirical Bayesian beamformer (EBB) approaches. Furthermore, we sought to infer the microstructure giving rise to specific functional activation patterns by comparing the myelin-informed model which was used to generate the activation with a set of test forward models incorporating different myelination patterns. We found that the direction of myelin changes, however not their magnitude, can be inferred by Bayesian model comparison. Finally, we apply myelin-informed forward models to MEG data from a visuo-motor experiment. We demonstrate improved source reconstruction accuracy using myelin estimates from a quantitative longitudinal relaxation (R1) map and discuss the limitations of our approach. Highlights We use quantitative MRI to implement myelin-informed forward models for M/EEG Local myelin density was modelled by adapting the local leadfields Myelin-informed forward models can improve source reconstruction accuracy We can infer the directionality of myelination patterns, but not their strength We apply our approach to MEG data from a visuo-motor experimentKeywords
This publication has 73 references indexed in Scilit:
- Relating quantitative 7T MRI across cortical depths to cytoarchitectonics, gene expression and connectomicsHuman Brain Mapping, 2021
- High‐precision magnetoencephalography for reconstructing amygdalar and hippocampal oscillations during prediction of safety and threatHuman Brain Mapping, 2019
- Reproducibility of T2* mapping in the human cerebral cortex in vivo at 7 tesla MRIJournal of Magnetic Resonance Imaging, 2014
- A general linear relaxometry model of R1 using imaging dataMagnetic Resonance in Medicine, 2014
- On the accuracy of T1 mapping: Searching for common groundMagnetic Resonance in Medicine, 2014
- Optimization and validation of methods for mapping of the radiofrequency transmit field at 3TMagnetic Resonance in Medicine, 2010
- A Parametric Empirical Bayesian framework for fMRI‐constrained MEG/EEG source reconstructionHuman Brain Mapping, 2010
- Increased SNR and reduced distortions by averaging multiple gradient echo signals in 3D FLASH imaging of the human brain at 3TJournal of Magnetic Resonance Imaging, 2008
- High‐resolution maps of magnetization transfer with inherent correction for RF inhomogeneity and T1 relaxation obtained from 3D FLASH MRIMagnetic Resonance in Medicine, 2008
- Interpreting magnetic fields of the brain: minimum norm estimatesMedical & Biological Engineering & Computing, 1994