Correcting for head movements in MEG inverse problem

To model the sources of measured magnetic fields, the relationship of the magnetic sensors to the head must be known [2]. This is usually accomplished by feeding currents through coils attached to the surface of the head [4]. The locations of the coils with respect to the sensor array are then computed on the basis of the measured signals. When evoked responses are studied, several responses usually need to be averaged. Usually the position of the head is assumed to be fixed during the experiment, but if the head has moved or data from different measurement sessions are combined, the combination of different head positions blurs the data spatially and adds a location bias. Therefore, the movements should be either eliminated or taken into account in the analysis. Subjects are typically asked to keep their head still during the measurements. This may be difficult even for cooperative subjects if the measurement is long or subject’s task includes movements. Using individual bite-bars helps avoiding head movements [6], but some subjects find this rather uncomfortable and it can not be applied in experiments requiring verbal responses. If the effects of measured head movements can be corrected for during data analysis, measuring MEG of less cooperative subjects, such as children, will be easier and there will be no additional restrictions on subject’s tasks. In this study we compare three different approaches for this kind of head movement correction.