Abstracts

Rationale: Often clinical magnetoencephalographic (MEG) recordings, especially recordings of ictal activity in patients with epilepsy, are accompanied with unstable head position and therefore the accuracy of source localization can be impaired. To recalculate the signal recorded from the head with unstable position, a novel approach – movement compensation (MC) [Uutela et al., NeuroImage 14: 1424-1431, 2001] – was developed based on one of two artifact suppression methods: the signal space separation (SSS) [Taulu et al., IEEE Transactions of Signal Processing, 2005; 53 (9), 3359-3372.] or the temporal signal space separation (tSSS) [Taulu and Simola. Phys Med Biol, 2006; 1759-1768; 51, 1759-17]. SSS rejects magnetic signals from outside the sensor array, and tSSS rejects the fields with synchronous time behavior inside and outside the sensor array. tSSS can be especially useful in suppression of artifacts that originate near the sensors. How these recalculations may affect the original brain signals and artifacts, has not been clearly documented, and we performed the study to answer this question.Methods: Somatosensory evoked MEG responses to electrical median nerve stimulation (Digitimer Ltd) were recorded with planar gradiometers and magnetometers (Elekta Neuromag Oy) while the subject’s head was kept in different positions and at least partly inside the MEG sensor helmet. The head position was continuously monitored by the recording system via four coils attached to the subject’s scalp. The source of the 20-ms response was determined with a single dipole model. We compared the source localization error, noise, goodness of fit, and confidence volume on data processed by MC-SSS vs. MC-tSSS. Results: Up to 5-cm shift in head position, MC-SSS increased noise in gradiometers (p=0.004); while tSSS significantly reduced the noise in gradiometers (p=0.0030) and magnetometers (p=0.0030), as well as reduced the mean localization error (4 cm for unprocessed data, 2 cm MC_SSS, and 1 cm MC_tSSS; p=0.0237) and increased the goodness of fit (p<0.0001). We defined two patterns of MC-enhanced disturbances: stimulus artifact increase which can be treated by tSSS, and random noise increase in very low head positions (5-6 cm shift). Conclusions: MC-tSSS significantly improves the source location as long as the source areas remain inside the sensor helmet. For sources close to the hand area of the somatosensory cortex, tSSS-based MC appears safe at least up to 3 cm head shift.