Abstracts

Rationale: Functional mapping of the eloquent cortex of pediatric patients undergoing neurosurgery is challenging with many children unable to cooperate with non-invasive procedures and having lower success rates with cortical stimulation and Wada procedures (Arya 2015, Epilepsy Res 110:78-87). TMS is a newer non-invasive mapping tool that is safe and well-tolerated by children (Narayana 2015, J Child Neurol 30(9):1111-24). While shown to be accurate in motor mapping (Narayana 2015, Pediatr Neurol 52(1):94-103), TMS-derived language maps have yet to be validated in this cohort. This study aimed to examine concordance between TMS and fMRI derived hemispheric dominance (HD) and localization of language specific cortices (LSC) in a clinical pediatric cohort. Methods: Children (n=86, 43 M) underwent functional brain mapping with TMS and fMRI at our institution. Language mapping with fMRI was done as patients performed object naming (ON_MRI) and verb generation (VG_MRI) covertly. Z-score images of activation were thresholded at p=0.005 with a cluster size of 500 mm³ to balance validity and reliability (Wilson 2017, Neuroimage Clin 16:399-408). The volume of activations in LSC that included frontal and temporal Brodmann areas (BAs) associated with language were tabulated for each task. Language mapping with TMS was done while patients performed overt object naming (ON_TMS). Five Hz TMS was delivered to the temporal and frontal cortices. TMS elicited speech errors (speech arrest, semantic error, or hesitation) were noted and their locations recorded. fMRI and TMS laterality indices were derived and between-method concordance was examined by deriving statistical performance metrics by hemisphere and by individual BAs. Results: LSC were identified by both tasks in fMRI, with ON_MRI identifying more LSC in right hemisphere (Figure 1). Consistent with ON_MRI, ON_TMS elicited speech errors when bilateral frontal and temporal regions were stimulated (Figure 1). While the rate of HD estimated by VG_MRI (82% left, 12% right, 6% bilateral) in this pediatric patient cohort was consistent with previous reports (Table 1b, Tzourio-Mazoyer 2017, Cortex 86:314-339), ON_MRI overestimated the incidence of right HD (23%). TMS underestimated left HD (62%) and overestimated bilateral HD (23%) (Table 1a). When compared to ON_MRI, ON_TMS had 75% sensitivity and 67% specificity, with an odds ratio of 6 (Table 1b). Of the eleven LSCs examined, BA 6, 44 and 9 (frontal) and BA 22 and 40 (temporal) were most commonly identified by both fMRI and TMS. The overall accuracy of ON_TMS across all LSC in both hemispheres was 73% (Table 1c), mainly due to high specificity, and the greatest contribution to the performance of TMS was from its effect on left hemisphere indicating that the right hemisphere (Table 1d, e) speech disruptions identified by ON_TMS were not reliable. Conclusions: This large-scale study in a pediatric patient cohort for the first time demonstrates good concordance of HD for language determined by TMS with that derived from fMRI. However, TMS derived HD does not capture the discordance of using different tasks. That the localization of LSC by TMS demonstrated only moderate concordance with fMRI reveals the need to use concordant tasks between the modalities and further optimization of TMS parameters. Funding: Not applicable.