Mapping of cortical language function by functional magnetic resonance imaging and repetitive navigated transcranial magnetic stimulation in 40 healthy subjects

Background: Functional magnetic resonance imaging (fMRI) is considered to be the standard method regarding non-invasive language mapping. However, repetitive navigated transcranial magnetic stimulation (rTMS) gains increasing importance with respect to that purpose. However, comparisons between both methods are sparse.

Methods: We performed fMRI and rTMS language mapping of the left hemisphere in 40 healthy, right-handed subjects in combination with the tasks that are most commonly used in the neurosurgical context (fMRI: word-generation = WGEN task; rTMS: object-naming = ON task). Different rTMS error rate thresholds (ERTs) were calculated, and Cohen's kappa coefficient and the cortical parcellation system (CPS) were used for systematic comparison of the two techniques.

Results: Overall, mean kappa coefficients were low, revealing no distinct agreement. We found the highest agreement for both techniques when using the 2-out-of-3 rule (CPS region defined as language positive in terms of rTMS if at least 2 out of 3 stimulations led to a naming error). However, kappa for this threshold was only 0.24 (kappa of <0, 0.01-0.20, 0.21-0.40, 0.41-0.60, 0.61-0.80 and 0.81-0.99 indicate less than chance, slight, fair, moderate, substantial and almost perfect agreement, respectively).

Conclusions: Because of the inherent differences in the underlying physiology of fMRI and rTMS, the different tasks used and the impossibility of verifying the results via direct cortical stimulation (DCS) in the population of healthy volunteers, one must exercise caution in drawing conclusions about the relative usefulness of each technique for language mapping. Nevertheless, this study yields valuable insights into these two mapping techniques for the most common language tasks currently used in neurosurgical practice.