Abstracts

The external validity of non-invasive language mapping has been examined in studies where MEG localization was confirmed by either intracarotid amobarbital procedure and/or intraoperative cortical stimulation during epilepsy surgery. A previous study using normal subjects indicates that MEG estimates of cerebral dominance for language and localization of temporal lobe language area are reliable on two different occasions for the same subjects. However, in order for MEG language mapping to be used clinically, it is important to establish its ability to produce reliable measurements for individual subjects from different raters as well as on different occasions. This study examined the reliability of MEG language mapping through two different raters over two different occasions for the same epilepsy surgery candidates. Nine epilepsy surgery candidates preformed an auditory recognition task for words. The task was repeated during two different sessions on the same day. Three lists of forty words were presented in each session. Thirty of the words were used as targets and ten words as distractors. Target words were presented for study before each MEG recording session. Subjects were instructed to raise their index finger to indicate that they recognized target words. Focal neurophysiological activity producing detectable magnetic fields at the surface of the head was modeled at each 4-ms time slice using the single equivalent current dipole model. Two raters (R1 and R2) analyzed data independently, and results of each rater were summarized in four parameters. 1. Inter-rater reliability
a. Asymmetry Index [AI = (left hemisphere [ndash] right hemisphere)/(left hemisphere + right hemisphere)] for magnetic sources showed agreement between two raters (Pearson correlation coefficient [italic]r[/italic] = .984).
b. Average distance of localized magnetic sources between two raters was .86cm (SD =.48cm).
c. Average root mean square magnetic value (RMS) from R1 was 87fT, and 116fT from R2.
d. Average spatial volume that encompasses 95% probability of source localization from two raters was .66 (SD=.73) and .40cm³ (SD=.57) respectively.
2. Intra-rater reliability
a. Each rater[rsquo]s AI over two sessions for the same patient showed a high degree of agreement with [italic]r [/italic]= .99 and .95 respectively.
b. R1[rsquo]s average distance of magnetic sources between two separate sessions was .85 cm, and R2[rsquo]s average distance between two sessions was 1.0 cm.
c. Average RMS over two sessions was 85 and 92fT for R1, and 115 and 127fT for R2.
d. Confidence volumes for each rater over two sessions were .61 and .67cm³ for R1, and .20 and .54cm³ for R2 In all four parameters of magnetic sources associated with receptive language function, there was a high degree of agreement between raters as well as within raters. These results suggest that MEG is a promising tool for determination of the localization of temporal lobe language areas for epilepsy surgery candidates.