Abstract
Magnetoencephalography (MEG) enables non-invasive real time characterization of brain activity. However, convincing demonstrations of signal contributions from deeper sources such as the hippocampus remain controversial and are made difficult by its depth, structural complexity and proximity to neocortex. Here, we demonstrate a method for quantifying hippocampal engagement probabilistically using simulated hippocampal activity and realistic anatomical and electromagnetic source modelling. We construct two generative models, one which supports neuronal current flow on the cortical surface, and one which supports neuronal current flow on both the cortical and hippocampal surface. Using Bayesian model comparison, we then infer which of the two models provides a more likely explanation of the dataset at hand. We also carry out a set of control experiments to rule out bias, including simulating medial temporal lobe sources to assess the risk of falsely positive results, and adding different types of displacements to the hippocampal portion of the mesh to test for anatomical specificity of the results. In addition, we test the robustness of this inference by adding co-registration error and sensor level noise. We find that the model comparison framework is sensitive to hippocampal activity when co-registration error is <3 mm and the sensor-level signal-to-noise ratio (SNR) is >−20 dB. These levels of co-registration error and SNR can now be achieved empirically using recently developed subject-specific head-casts.
Highlights
Magnetoencephalography (MEG) is a non-invasive neuroimaging technique that measures electromagnetic brain activity with millisecond temporal resolution
The increase in spatial spread or complexity in the bottom row occurs because it requires more non-hippocampal sources to describe MEG data arising from a single hippocampal source than would be needed if the true source were modelled
With respect to single-simulation ΔFanatomical values corresponding to solutions shown in Fig. 4A, we find that for all three algorithms, the combined model has a higher Free energy value than the cortical model; single simulation ΔFanatomical Minimum Norm Estimate (MNE) = 1.4, Empirical Bayes Beamformer (EBB) = 10.6, Multiple Sparse Priors (MSP) = 73.2
Summary
Magnetoencephalography (MEG) is a non-invasive neuroimaging technique that measures electromagnetic brain activity with millisecond temporal resolution. In order to localise the spatial origin(s) of such activity, anatomical and electrophysiological information is used to constrain the solution space Whilst this general framework is wellestablished for neocortical sources (Gross et al, 2003; Hämäläinen et al, 1993; Henson et al, 2009; Lopes da Silva, 2013; Vrba and Robinson, 2001), reconstruction of deep sources remains controversial (Hämäläinen et al, 1993; Mikuni et al, 1997; Riggs et al, 2009; Stephen et al, 2005). Recent evidence suggests that the current source density generated by the hippocampal pyramidal cell layer is at least twice that of the neocortex, which might compensate to some degree for its distance to the sensors (Attal et al, 2012; Murakami and Okada, 2015, 2006)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
