Source localization with MEG evoked by the olfactory oddball paradigm

1 Introduction Cortical areas activated by olfactory stimulation have not been made clear yet, though a few studies were reported [1-3]. We are trying to localize the sources of olfactory MEG. The olfactory response becomes quickly weak because of subject's habituation. Then the data for source localization have little number of averaging and include much noise. The estimation error of localization is closely correlated to the signal to noise ratio (S/N). In order to estimate the sources accurately, we developed a noise reduction method using wavelet transform to do "time-frequency analysis"[4,5]. Adopting approximating functions with compact support, the wavelet analysis has advantages over traditional Fourier transform method, where the signal contains discontinuities and sharp spikes, such as evoked responses in MEG data. We applied this method to the olfactory MEG data. Four response signals were extracted at different latencies. The source equivalent current dipoles (ECDs) of each signal were estimated by the simulated annealing method (SA). The estimated locations and directions of the ECDs show that the source of olfactory MEG differs depending on their latencies and also the kind of odor stimuli. 2 Methods 2.1 Olfactory MEG Olfactory MEGs were measured with a whole cortex type Neuromag-122™ SQUID system in Osaka Life Electronics Research Center. Six healthy right-handed subjects (6 male, between 21-22 years old, with informed consents) participated in this experiment. An odor pulse is given in the nostril through a Teflon tube synchronously with subject's inhalation, but once in three or four consecutive inhalations. Duration of the stimuli is 300ms, a rising time below 20ms and delay from the onset of the valve at 100±2ms. Prepared odors were 1% isoamyl acetate, 1% isovaleric acid, 1% geraniol and odorless air, and two among them were used in one experiment. One of the two odors was randomly given more frequently than the other in the ratio of 3:1. Subjects were asked to count the number of occurrence of the rarer (target) odor. During measurements, the subjects hear white noise (SPL 70dB) through earphones, so the sounds from the valve do not interfere with the responses. For preventing eye movements, the subjects gazed at a point in front of them. MEG was recorded at a sampling rate of 400Hz, through a band-pass filter of 0.03-100Hz. Responses to the target (rare) and non-target (frequent) odors were separately averaged with respect to the onsets of stimulation. The number of averaging was about …