Abstracts

Rationale: Evaluation of language activity of the brain is important for planning epilepsy surgery. Magnetoencephalography (MEG), a non-invasive neuroimaging procedure, is frequently used for this purpose. We previously presented the usefulness of distributed source analysis of language MEG by using dynamic statistical parametric maps (dSPMs) (AES 2010). However, the dSPM results are highly affected by the noise covariance obtained from the background signals. The purpose of this study is to explore how the noise covariance affects the representation of language activity recorded on MEG. Methods: Four intractable epilepsy patients (male: 3, female: 2, age: 15-54) were retrospectively studied. MEG was recorded with a 306-channel whole-head system at a sampling rate of 1000Hz. In all patients, high-resolution 3T anatomical MRI data were acquired with magnetization-prepared rapid acquisition gradient-echo (MPRAGE). During MEG recording, patients performed a semantic word decision task with 160 English words presented visually. The interstimulus interval was 3000ms. Patients are asked to stair a blank screen for the first 30 seconds prior to the word presentation. Noise covariance was computed from (1) the interstimulus data between -500 and -10ms prior to each stimulus (cov-ISI) and (2) the consecutive raw data of the first 30 seconds without any stimuli (cov-raw). MEG data epochs from 500ms pre-stimulus and 1000ms post-stimulus period were averaged after preprocessing. We calculated dSPMs (Dale et al., 2000) by using cov-ISI and cov-raw separately. Laterality index (LI) was obtained by deploying regions of interest (ROIs) on the MRI-derived cortical surface, including superior temporal, middle temporal, supramarginal, inferior parietal cortices, opecular and triangular parts of the inferior frontal lobe on both hemispheres, based on a priori knowledge of Wernicke and Broca areas. LI was obtained by LI = (L-R)/(L+R), where L and R is the number of unit dipoles with an F value higher than the threshold in the ROIs within a time window of 250-550ms post stimuli. The threshold was determined as the half of the maximum value over all ROIs. Language laterality was estimated based on the LI as follows; 0.1:left, 0.1>LI>-0.1:bilateral, -0.1:right. Results: In visual inspection, dSPMs showed smaller activation by using cov-ISI than cov-raw outside the language-associated ROIs. The maximum activation value within the ROIs was larger when computed by using cov-ISI than cov-raw in four of five (80%) patients. The overall hemispheric dominancy was consistent with the laterality obtained by functional MRI in four patients, regardless of using cov-ISI or cov-raw. One patient showed left hemispheric dominancy and bilateral representation by using cov-ISI and by cov-raw, respectively.Conclusions: These results may suggest dSPMs represent more language-specific activity by using cov-raw than cov-ISI, although these covariances did not change the overall laterality in 80% of the patient. This study encourages further investigation in a larger patient group.