Source localization scale correction for Beamformer analysis

Magnetoencephalography measurements are often processed by using imaging algorithms such as beamforming. The estimated source magnitude tends to suffer from unbalanced scaling across different brain locations. Hence, when examining current estimates for source activity it is vital to rescale the estimated source magnitude, in order to obtain a uniformly scaled image.

We present a generalized scale correction method (N_empty) that uses empty room MEG measurements to evaluate the noise level.

The location bias and spatial resolution of the estimated signal indicated that some scaling correction needs to be applied. Of all the scale correction methods that were tested, the best correction was achieved when using N_empty.

We show that a diagonal matrix does not reflect the true nature of the noise covariance matrix. Hence, diagonal matrix based methods are sub-optimal.

We recommend adding empty room MEG measurements to each experimental recording session, for purposes of both scale correction and beamformer performance verification.