The localization of rhythmic brain activity in patients with brain tumors using magnetoencephalography

Multi-channel MEG data is used to localize electric brain activity in the head by constructing a mathematical inverse model that gives the relationship between the location and orientation of the source (a current dipole) and the measured magnetic fields. When this technique is used to localize the generators of spontaneous brain rhythms (like the alpha rhythm, 8-12 Hz or the delta rhythm, 1-4 Hz) one faces the problem that the non-linear inverse problem has to be solved on large amounts of data. In our study, the inverse problem was solved efficiently by disregarding samples with low amplitude and samples with large initial guess errors, by splitting up the inverse problem into a linear and a non-linear part and by using a global search algorithm based on pre-computed tables applied with each time sample. When this technique is applied on the delta band of MEG data sets of patients with brain tumors, clusters of good fitting dipoles are found at the tumor boundaries. The numerical efficiency of the applied algorithm is sufficient to be applied in clinical practice on a routine basis.