408 Electroencephalography and clinical Neurophysiology , 87 (1993) 408-416 © 1993 Elsevier Scientific Publishers Ireland, Ltd. 0013-4694/93/$06.00 EEG93548 A comparison of the localization of spontaneous neuromagnetic activity in the frequency and time domains * Claudia Tesche and Matti Kajola Low Temperature Laboratory, Helsinki UniL,ersity of Technology, SF 02150 Espoo (Finland) (Accepted for publication: 30 July 1993) Summary We investigated the localization of current sources for spontaneous magnetoencephalographic data in the frequency and time domains. The two analysis techniques yielded complementary information about the underlying neuronal generators. Phase-coherent sources for occipital 10 Hz alpha band activity were identified in the frequency domain from the dependence of the equivalent current dipole strengths on the phase of the Fourier transform. Source localization in both the frequency and time domains was used to analyze data containing interictal spike and slow-wave activity. Predominantly 2-6 Hz spectral components localized in the frequency domain were found within an 8 cm 3 volume centered at the time-domain source location of the spike. In general, the characteristics of the noise sources in magnetoencephalographic systems favor the use of frequency-domain analysis for rhythmic spontaneous activity. Key words: Spontaneous activity; FF]? dipole source localization; Brain mapping; Magnetoencephalography; Electroencephalography; (Human) Electroencephalographic (EEG) and magnetoen- cephalographic (MEG) data are recorded as time-do- main signals which have been bandpass filtered and sampled at some convenient rate. The data can be transformed from the time domain into the frequency domain by taking the fast Fourier transform (FFT) of each channel of data for some epoch. This procedure preserves useful information about the spatial distribu- tion of the underlying neuronal generators (Lehmann and Michel 1989; Liitkenh6ner 1992). Thus, source localization algorithms may be applied to data in both the time and frequency domains (Liitkenh6ner 1992). The utility of these methods depends not only on the characteristics of the neuronal networks which are responsible for the signal of interest and the back- ground brain activity, but also on the sources of noise in the measurement system. Although source localization in both MEG and EEG requires a model for the brain as a conducting volume, source localization algorithms for MEG data do not require a model for the skull and are thus more conve- nient than those for EEG data for the illustration of novel features of frequency-domain analysis (for a re- view, see H~im~il~iinen et al. 1993). We have analyzed MEG data in the frequency domain for spontaneous activity in 10 normal subjects and 17 patients. Localiza- tion of frequency-domain current dipole sources was possible in all data sets. In this paper, we illustrate differences between frequency- and time-domain local- izations of alpha-band activity with data from a normal subject during relaxed wakefulness. The advantage of combining both frequency- and time-domain analyses in data sets containing both transient and oscillatory activity is illustrated with data containing spike and slow-wave activity from a child with Landau-Kleffner syndrome. Finally, we discuss the impact of MEG system noise on frequency- and time-domain localiza- tion. Correspondence to: C. Tesche, Low Temperature Laboratory, Helsinki University of Technology, SF 02150 Espoo (Finland). Fax: + 3580-451-2969. * This work was supported by the Academy of Finland and the Sigrid Jus61ius Foundation. Theory The relationship between frequency-domain and time-domain source localizations can be seen by exam- ining the dependence of the EEG or MEG signals on the current source distribution. In general, any current