What and where in electromagnetic brain imaging
To understand the brain, we need to observe both the nature and dynamics of its activity (the "what"), and the location or distribution of its sources (the "where"). This paper proposes a new approach based on standard data-driven linear analysis, in which these two elements are derived in parallel from separate columns of the analysis matrix. A subset of columns enhances the activity of interest, and its complement defines spatial filters that suppress that activity (null filters). Each null filter is combined with an anatomy-dependent source model to estimate a subset of the sources space (zero set) in which the source might reside, and the final estimate is derived from the intersection of zero sets. There is no theoretical limit to the accuracy with which the location of a source can be estimated, but practical limits may arise from noise in the data, imperfect calibration, or an incomplete or inaccurate source model.