Preoperative transcranial magnetic stimulation for picture naming is reliable in mapping segments of the arcuate fasciculus

In preoperative planning for neurosurgery, both anatomical (diffusion imaging tractography) and functional tools (MR-navigated transcranial magnetic stimulation) are increasingly used to identify and preserve eloquent language structures specific to individuals. Using these tools in healthy adults shows that speech production errors occur mainly in perisylvian cortical sites that correspond to subject-specific terminations of the major language pathway, the arcuate fasciculus. It is not clear whether this correspondence remains in oncological patients with altered tissue. We studied a heterogeneous cohort of 30 patients (fourteen male, mean age 44), undergoing a first or second surgery for a left hemisphere brain tumour in a language-eloquent region, to test whether speech production errors induced by preoperative transcranial magnetic stimulation had consistent anatomical correspondence to the arcuate fasciculus. We used navigated repetitive transcranial magnetic stimulation during picture naming and recorded different perisylvian sites where transient interference to speech production occurred. Spherical deconvolution diffusion imaging tractography was performed to map the direct fronto-temporal and indirect (fronto-parietal and parieto-temporal) segments of the arcuate fasciculus in each patient. Speech production errors were reported in all patients when stimulating the frontal lobe, and in over 90% of patients in the parietal lobe. Errors were less frequent in the temporal lobe (54%). In all patients, at least one error site corresponded to a termination of the arcuate fasciculus, particularly in the frontal and parietal lobes, despite distorted anatomy due to a lesion and/or previous resection. Our results indicate that there is strong correspondence between terminations of the arcuate fasciculus and speech errors. This indicates that white matter anatomy may be a robust marker for identifying functionally eloquent cortex, particularly in the frontal and parietal lobe. This knowledge may improve targets for preoperative mapping of language in the neurosurgical setting.