Abstracts

Rationale: Prior to resective brain surgery it is important to know that areas to be resected do not carry important function. The gold standard for clearing brain areas for resection involves invasive electrical stimulation mapping (ESM). Functional magnetic resonance imaging (fMRI) may be a tool to help perform this noninvasively. To our knowledge this approach has not been formally tested on its validity to predict ESM results, particularly on its potential clinical utility as a tool to clear brain areas for function. We investigate the spatial correlation of ESM results with the blood oxygen level-dependent (BOLD) response for language and motor tasks. This study aims to define the role of fMRI to delineate areas that should be spared from resection. Methods: fMRI was recorded while patients were doing a finger tapping task (N=8) or a picture naming task (N=5). ESM was performed with analogous tasks for language mapping. Post-implantation CT was coregistered to the pre-implantation fMRI using an automated method that accounted for brain shift and the craniotomy. With this coregistration, the BOLD response in the vicinity of electrodes was correlated with the results of ESM. 262 electrodes sites were investigated for language function and 473 for motor function. Sensitivity, specificity and predictive values were calculated for validation of fMRI. We used different region of interest volumes (ROIV) under the electrodes to maximize negative predictive value (NPV) and examined the correlation of expressive and receptive language functions separately. Results: Significantly higher t-scores were associated with ESM positive electrodes compared to ESM negative electrodes for both, motor (5.9 vs. 3.0; p<0.01) and language (2.3 vs. 1.6; p<0.01) functions. At a t-score threshold of 2.5, and a ROIV of 16mm the sensitivity and negative predictive values are both above 99%, which is excellent for clinical use. Using the same parameters language showed NPV that would not be sufficient for clinical utility (< 85%). Including only expressive language areas the NPV increased to a clinical useful value of 99% and the sensitivity to 97%. Conclusions: fMRI may localize expressive language function with the same accuracy as motor function, which reaches a clinically useful degree. Using conjunct analysis of several language tasks as well as additional post-hoc analysis might provide better results for global analysis of language.