Robustness of cortical distortion metrics to unmodelled OPM errors
In the authors' words
The estimate of cortical current flow from magnetic fields measured outside the head is an ill-posed problem. Recent work using head casts to minimise co-registration errors in conventional cryogenic magnetoencephalography (MEG) has shown that ground truth anatomy can be used to judge between different inversion schemes and metrics. Inversion algorithms or metrics that are insensitive to anatomical distortion are unlikely to reflect correct assumptions about the underlying current distribution. Here we aim to identify constraints when using this approach to validate current flow estimated from recordings made with optically pumped MEG (OP-MEG). To do so, we used simulations contaminated by different sources of (typically unmodelled) error. For example: error in the estimate of the true sensor positions or orientations and the effects of unmodelled gain changes. In accordance with previous work, with no unmodelled errors, distortion was minimized when the simulation and the inversion assumptions matched. We then quantified the amount of error (in mm) relative to the true cortical mesh that one might expect in empirical recordings (given gain, position and orientation errors). We found that the approach was tolerant to independent sensor position errors of less than 2 mm (per axis); orientation errors of less than 5 degrees; fixed array displacements of less than 10 degrees and to random sensor gain errors of less than 10%. Additionally, we tested this approach using two empirical OPM datasets. We found that source reconstructions based on both minimum norm and beamformer assumptions were sensitive to the underlying anatomy.
Appeared: Thursday, November 19. bioRxiv. Preprint, not yet peer-reviewed.