In the present study, we estimated ictal onset zones in five patients with intractable frontal lobe epilepsy by using both dSPMs and a single-dipole model. In all spikes obtained from the early ictal discharges, the dSPM localized activity consistent with the clinically-determined ictal onset zone at a lobar level, except for one spike in Patient 2. The dSPM activation in this spike showed larger activation areas, including the clinically-determined ictal onset zone. Moreover, dSPMs had more restricted activation in these ictal spikes than interictal spikes. It is well known that the irritative zone obtained from the interictal spikes is often larger than the ictal onset zone (Lüders and Awad, 1991
). DSPM findings of this study are in agreement with this phenomenon.
Since the single-dipole analysis always has the difference between measured and forward-calculated magnetic fields, many ictal MEG studies have applied a threshold of GOF values for selecting ECDs as appropriate sources (Oishi et al., 2002
; Tanaka et al., 2004
; Yoshinaga et al., 2004
; Tanaka et al., 2005
). The threshold has been set at 80-90% in these studies. This is a form of appropriate bias for increasing the likelihood of single-dipole analysis, when many ECDs are available after eliminating ECDs with a low GOF value. Several MEG studies have selected the channels of MEG to be used in the single-dipole analysis for obtaining better GOF values (Bast et al., 2004
). This is another bias which may be useful only when the selection is adequate for representing the epileptic activity. In this study, we used all the MEG channels for single-dipole analysis. The number of ECDs decreased as the threshold of goodness-of-fit (GOF) values increased. No ECDs or only one ECD with the threshold of GOF > 80% in each case. The locations of some ECDs obtained with the thresholds of GOF > 60% and 70% were inconsistent with the clinically-determined ictal onset zones. These results may suggest the usefulness of ECDs may be limited in the evaluation of these ictal MEG spikes. Theoretically, dSPMs can represent all the magnetic fields by applying a source distribution model, assuming that the sources are distributed in the large patch of cortex (Dale and Sereno, 1993
; Dale et al., 2000
). Therefore, thresholds such as GOF values are considered unnecessary for dSPM analysis. On the other hand, determination of the appropriate amount of activated cortex is an important issue in distributed source models including dSPM. In dSPM, this issue is highly dependent on F values for statistical mapping. Smaller F values would show a larger area of activation, and larger F values would show no activation. In any case, the inverse solution used for making dSPMs explains all magnetic fields. We applied F values for representing the strongest activation, based on the assumption that epileptic activities are larger than other background activities. In addition, the activation areas were consistent at a lobar level under several different F values in all patients. Further study will be needed for investigating the accuracy of dSPM at a sublobar level by comparing dSPMs and the spike involvement on intracranial EEG.
Source localization of ictal discharges becomes difficult after seizure activity propagates (Tanaka et al., 2004
), thus early ictal spikes may be more accurate for estimating the seizure onset zone (Tilz et al., 2002
; Assaf et al., 2003
). However, the low signal-to-noise ratio in the early spikes may complicate obtaining appropriate sources using single-dipole modeling. This may explain the low GOF values and widespread distribution of ECDs in the present study. In contrast, dSPMs showed a localized activation which was reproducible between the different ictal spikes, suggesting that they may be useful for analyzing these spikes with the low signal-to-noise ratio.
The small number of patients and the lack of a confident ictal onset zone determined by intracranial EEG may cause a limitation to the results of this study. However, we consider this study is still valuable for improving MEG analysis since this is the first report in which dSPMs are applied for analyzing ictal data. Two patients underwent surgery with a favorable outcome, removing the area of cortex which was consistent with the results of dSPMs. Further investigation in a larger patient group with intracranial EEG will provide an improvement on these issues.
Overall, the results of our patients suggested that dSPM provided superior source localization of early ictal spikes compared to single-dipole source analysis. By using dSPM, ictal MEG may be more informative than the single-dipole analysis alone in presurgical evaluation of epilepsy.