Validation of head movement correction and spatiotemporal signal space separation in magnetoencephalography
Highlights
► We evaluated novel methods for artifact suppression and movement correction in MEG data recorded in 20 subjects with controlled magnetic artifacts and head movements. ► The results show that methods based on signal space separation can reliably suppress interference from nearby sources. ► Head movement correction recovered evoked responses of satisfactory quality in all test conditions.
Introduction
Magnetoencephalography (MEG) reveals the dynamics of cortical processing with millisecond resolution (Hämäläinen et al., 1993). It has clinical applications in pre-operative planning and localization of epileptic foci (Mäkelä et al., 2006).
Significant head movements and magnetized impurities near or on the head have traditionally been regarded as exclusion criteria for MEG recordings. The signal space separation (SSS) method (Taulu and Kajola, 2005) provides a novel approach for suppressing interference originating outside of the MEG sensor array. In addition, the SSS method can be used for head movement correction and standardization of the head position (Taulu et al., 2005).
The temporal extension of SSS, tSSS (Taulu and Simola, 2006) further improves the shielding capabilities of SSS. The tSSS method is able to suppress interference from sources very close to the sensors, e.g. from magnetic particles on the head. The method has been experimentally demonstrated by analyzing unaveraged, tSSS-processed auditory responses from a healthy volunteer with controlled interference sources, though without head movement (Taulu and Hari, 2009). Recently, tSSS has also been evaluated in clinical studies involving epileptic patients with vagus nerve stimulators (Carrette et al., 2011) and Parkinsonian patients implanted with deep brain stimulators (Airaksinen et al., 2011). Using tSSS simultaneously with head movement correction can improve data quality further. For example, Medvedovsky et al. (2007) studied averaged somatosensory responses of one healthy volunteer from a reference head position and from five distinctly differing deviant positions, i.e. the head was initially in the reference position and after few seconds shifted to a deviant position for the rest of the recording. Movement correction improved the mean localization error of magnetic N20 responses from 39 to 21 mm, while combined tSSS and movement correction improved the mean error further to 9 mm.
Another test of SSS-based movement correction was a study involving a group of 19 children of 8–12 years of age performing a cognitive task (Wehner et al., 2008). When movement correction was applied, the goodness-of-fit improved slightly, and the confidence volume of a dipole cluster shrunk by 19%. The efficiency of the SSS-based head position transformation was evaluated by (Lioumis et al., 2007).
In the present study, our aim was to systematically assess the reliability of tSSS and movement correction. To this end, we recorded MEG data from 20 volunteers in four different controlled conditions. We also recorded signals from a phantom subjected to similar conditions. The data were processed with tSSS for suppression of artifacts and with an SSS-based movement correction. The results were compared to a reference recording where the head was held as stationary as possible. The conditions included discrete movements of the head, continuous regular movement, and continuous movement with magnetic impurities on the head.
Discrete head position shifts are typical in all MEG experiments where the subject is required to keep the head still for several minutes or even longer. Such movements are difficult to detect without continuous head position tracking. Some subjects and patients are not willing or able to maintain a steady head position and may continuously move their heads. Reliable analysis of such recordings requires effective movement correction and interference suppression.
Section snippets
MEG system and head position tracking
All MEG recordings were performed with an Elekta Neuromag® system (Elekta Oy, Helsinki, Finland), which samples the magnetic field distribution by 510 coils at distinct locations directly above the cortex. The coils are configured into 306 independent channels consisting of 102 magnetometers and 204 planar gradiometers.
The location of the subject’s head relative to the MEG sensors was determined with the aid of four head position indicator (HPI) coils attached to the scalp. First, three
Phantom data
The results of phantom dipole localization are presented in Table 1 and in Fig. 1. With tSSS and MC, the dipole signals could be reliably reconstructed. For rotation speed less than 5 deg/s, single epochs were clear in the raw data and could be localized accurately. In faster rotations the finite window length (200 ms) employed in the HPI amplitude estimation reduced the accuracy of movement tracking. Still, after averaging the dipole localizations were 3 mm or less from the known locations.
Summary
After tSSS and movement correction, evoked responses sufficiently good for source modeling could be recovered in all conditions. The combined tSSS and MC method showed significantly smaller deviations in source localization than tSSS alone.
With tSSS and MC, the average localization difference compared to the reference data was reduced to the level of the intersite differences, i.e. to about 5–8 mm for most responses. The AEF conditions with continuous head movements displayed more variability
Conclusion
We have demonstrated the reliability of temporal signal-space separation and head movement correction in MEG evoked-response measurements. These methods are especially important in clinical settings, where patients may have sources of strong magnetic interference close to their heads (e.g. a brain stimulator) or they may perform involuntary movements, as in Parkinson’s disease and epilepsy. Our data also show that magnetic auditory evoked responses are more affected by continuous movement than
Acknowledgements
Lauri Parkkonen and Veikko Jousmäki were supported by the Academy of Finland (National Centers of Excellence Programme 2006–2011). Additionally, Lauri Parkkonen was supported by the aivoAALTO research project of Aalto University and Veikko Jousmäki by the European Research Council (Advanced Grant #232946). Jussi Nurminen was supported by Jenny and Antti Wihuri Foundation and the Instrumentarium Foundation.
References (25)
- et al.
Combined EEG and MEG analysis of early somatosensory evoked activity in children and adolescents with focal epilepsies
Clin Neurophysiol
(2007) - et al.
Magnetoencephalography in epilepsy patients carrying a vagus nerve stimulator
Epilepsy Res
(2011) - et al.
Integrating sensory and motor mapping in a comprehensive MEG protocol: clinical validity and replicability
NeuroImage
(2004) - et al.
Recording epileptic activity with MEG in a light-weight magnetically shielded room
Epilepsy Res
(2008) - et al.
Intrasubject reliability and validity of somatosensory source localization using a large array biomagnetometer
Electroencephalogr Clin Neurophysiol
(1994) - et al.
A brush stimulator for functional brain imaging
Clin Neurophysiol
(2007) - et al.
Standardization of MEG sensors by the signal space separation method
Int Congr Ser
(2007) - et al.
Head movements of children in MEG: quantification, effects on source estimation, and compensation
NeuroImage
(2008) - et al.
Replicability of MEG and EEG measures of the auditory N1/N1m-response
Electroencephalogr Clin Neurophysiol
(1998) - et al.
Effects of DBS on auditory and somatosensory processing in Parkinson’s disease
Hum Brain Mapp
(2011)
Neuromagnetic mapping of brain function
Radiology
MEG in the study of higher cortical functions
Cited by (51)
The hidden brain-state dynamics of tACS aftereffects
2022, NeuroImageMagnetoencephalography with optically pumped magnetometers (OPM-MEG): the next generation of functional neuroimaging
2022, Trends in NeurosciencesMulti-modal imaging of the auditory-larynx motor network for voicing perception
2022, NeuroImageCitation Excerpt :MNE-C and MNE-python (https://mne.tools/stable/index.html) (Gramfort et al., 2013, 2014) were used for MEG data analysis. First of all, all data sets were filtered using Maxfilter to eliminate noise originating from outside the brain (Nenonen et al., 2012). Next, the MEG signals were digitally filtered from 1 to 30 Hz using a bandpass Butterworth filter.
Non-stationarity Removal Techniques in MEG Data: A Review
2022, Procedia Computer Science