Abstracts

Rationale: TMS is increasingly used to non-invasively localize critical language cortices in children undergoing surgery for refractory epilepsy or brain tumor, but the optimal language mapping protocols are yet to be defined. The aim of this multicenter study was to evaluate existing TMS language mapping studies. Our hypothesis was that this information would then allow standardization of pediatric language mapping protocols across institutions with the ultimate goal of optimizing TMS language mapping parameters and improving accuracy.

Methods: Patients who had undergone clinical TMS language mapping were identified at the participating institutions by a retrospective chart review and their demographic, clinical, and TMS parameters were extracted into a data repository.

Results: We identified 315 patients (79% ≤ 18 y; 159 F) with refractory epilepsy or brain tumor in whom presurgical TMS language mapping was attempted. The majority of patients were on anti-seizure medications. Language areas were localized by repetitive TMS that generated a ‘virtual lesion’ in the stimulated area. Repetitive TMS trains were timed to correspond to image presentation during a graphic or color naming task. Speech errors were scored by post hoc video review. Both hemispheres were assayed in 294 (93%) patients, with 84% of those considered complete with bilateral frontal and temporal lobe mapping (Table 1). As needed, TMS intensity was reduced to accommodate patient pain/discomfort. Only 6 (2%) patients could not complete mapping due to pain. An average of 183 trains of TMS were applied at 5 Hz with an average intensity of 36.2% machine output (MO) (Table 2). TMS intensity during language mapping was compared to the resting motor threshold (rMT) in the corresponding hemisphere for each patient. Averaged over all participants, TMS intensity used for language mapping was 72% of rMT in the left frontal, 76% of rMT in left temporal, 71% in right frontal, and 77% of rMT in right temporal regions. The types of speech errors elicited included speech arrest, semantic errors and slowing or hesitation (performance errors) (Table 2). Based on the type and number of errors, hemispheric dominance (HD) for language was estimated in 86% of the patients in whom both hemispheres were mapped, with 62% deemed left HD, 13% right HD, and 25% bilateral HD.

Conclusions: This multicenter data indicates that meaningful language mapping data can be derived in children with refractory epilepsy or tumor by 5 Hz TMS at intensities lower than the rMT (~74%). These results indicate that the rMT is not a good predictor of language threshold at least in this cohort. From a practical perspective, our findings indicate that language mapping is feasible in cohorts with high rMT such as children and patients taking anti-seizure medications.

Funding: Please list any funding that was received in support of this abstract.: Funded in part by the infrastructure grant by AES.