HEAD MOTION EVALUATION AND CORRECTION IN MAGNETOENCEPHALOGRAPHY

Head motions during magnetoencephalography (MEG) data acquisition lead to inaccuracy in MEG signal localization and statistical sensitivity. Multiple head motion correction methods have been developed and validated to insure the head motion effects are removed, with the aim of improving localization accuracy and statistical sensitivity. This study investigated the amount and extent of the head motion during MEG resting state recordings with 80 subjects and supported the previously known downward motion of the head during scanning. Rotational motion was quite negligible and had no preferred direction. These findings led to investigation of the effectiveness of two motion correction methods, Single Space Separation (SSS) and General Linear Modeling (GLM) for correction of translational motion. SSS and GLM were evaluated and compared by five assessment criteria: Percent Root Difference (PRD), Pearson Product-Moment Correlation Coefficient (CC), Signal-to-Noise Ratio (SNR), localization accuracy and signal coherence. Quantitative comparison revealed that SSS is superior for data accuracy, resemblance and localization precision when applied to the pre-filtered recordings compared to GLM. The localization accuracy of SSS is within 1-3 mm for all motion directions up to 2 cm. SSS reduces the coherence strength and the total number of the coherent links in the pre-filtered data. GLM improves/reduces data accuracy and resemblance when applied to unfiltered/pre-filtered data respectively. The localization precision of GLM reduces approximately linearly with the motion extent. GLM improves localization accuracy in the unfiltered/pre-filtered recordings by a factor of 2-4. GLM does not change coherence strength and only slightly decreases/increased the total number of the coherent links on the sensors’ level when applied to the unfiltered/filtered data respectively. This study concluded that the SSS method yields better localization accuracy and better data quality improvement than GLM when applied to the pre-filtered MEG recordings. SSS was further investigated here and a route for its improvement was suggested. This route uses oblate/prolate spheroidal harmonics in place of the spherical harmonics expansion and preserves the simplicity of the translation and rotation transformations of the spherical harmonic functions. The preservation is achieved via forward and inverse spherical to spheroidal expansion coefficients transformations allowing motion correction on spherical harmonics coefficients.

Grantor dc:publisher

Wake Forest University

Year dc:date.issued

2015

Author dc:creator

• McGowin, Inna

• coherence

Language dc:language.iso

en

Handle dc:identifier.uri

http://hdl.handle.net/10339/57118

OAI identifier oai:identifier

oai:wakespace.lib.wfu.edu:10339/57118