Basic limitations of linear inverse solutions: a case study

This paper enumerates and illustrates some limitations of distributed linear inverse solutions applied to the reconstruction of the sources of the brain electric activity. Selected computer simulated images illustrating these drawbacks for the case of the LORETA solution are presented. It is shown how limitations such as existence of spurious sources, the high level of blurring and the incorrect estimation of the source strength can distort the images reconstructed with linear solutions even for single sources and perfectly accurate data. The problem of why measures designed to evaluate the performance of dipole localization methods should not be applied to the evaluation of distributed solutions is discussed. Alternatives to extract useful information about brain functioning in spite of the limitations of these methods are mentioned.