Minimum-norm estimate (MNE)

The minimum-norm estimate (MNE) resolves the inverse problem's non-uniqueness by choosing, among all source distributions that fit the data, the one with the smallest total power (smallest L2 norm). It has a closed-form linear solution: the estimated sources are a regularized, lead-field-weighted projection of the data, with a Tikhonov term balancing data fit against source energy.

MNE is stable, fast and assumption-light, but the minimum-energy prior has a well-known bias: it favours superficial sources, because deep sources need large amplitudes (and thus large norm) to explain the same sensor signal. This depth bias is why raw MNE tends to place activity on gyral crowns near the sensors. Depth weighting (rescaling the columns of the lead field) and noise normalization (dSPM, sLORETA) were developed specifically to counter it.