Abstracts

Rationale: fMRI and intracranial electrocorticography (ECoG) are being increasingly used to correlate and elaborate the results of invasive electrocortical stimulation (ESM) language mapping. An accurate assessment of the correlation of results obtained by less invasive methodologies such as fMRI and ECoG with the more invasive ESM requires accurate coregistration techniques. These methods have not yet been well described nor has there been significant data presented on patients who have undergone all three functional mapping techniques utilizing the same tasks and stimuli for individual subjects. We present data to further investigate the clinical utility of fMRI and ECoG as compared to gold standard of ESM. Methods: A 3D MRI image is obtained during the pre-op fMRI session. Following intracranial electrode implantation, an MRI and CT are acquired extending from top of the skull to cervical spine allowing for maximally accurate rigid body cross-modal registration. The CT is thresholded to reveal electrodes as clusters of bright voxels (www.bioimagesuite.org) and the centroids are used to define electrode position. A rigid registration between the CT and post-op MR enables electrode projection onto the aligned and stripped post-op MR. Surface reconstructions of the post-op MR can contain distortion due to presence of electrodes, thus a nonlinear intensity based registration is performed to align the stripped pre-op and post-op MR scans. This registration accounts for brain displacement and allows for accurate overlay on a brain surface with clear anatomic landmarks. fMRI activation maps (False Discovery Rate = .05) using both a visual naming and reading task are overlayed on this same pre-op MR. ECoG (event-related responses in gamma frequencies) and ESM are performed utilizing the same task design and stimuli used for the fMRI. Results: We find accurate electrode representation on individual patient brains with these alignment techniques allowing for anatomic localization and comparison of the results of fMRI, ECoG and ESM. Results of sensitivity and specificity (n=9) between ESM and fMRI vary between 33-53% and 81-97% respectively. Between ESM and ECoG results show between 14-53% and 71-89% respectively. Comparisons between fMRI and ECoG are similar. Results do not differ significantly between language tasks. Results do change with different fMRI thresholds, but sensitivity does not increase to the level of clinical utility. Given the relatively high specificity between these modalities both fMRI and ECoG might be used to create preliminary functional maps to guide ESM. Conclusions: The current registration techniques can be used to accurately study the relationship between various language mapping modalities. The current data do not support the clinical utility of these techniques to determine critical language sites. The reason for the discrepancies between these modalities is unclear. However, the varying mechanisms by which these modalities (activation v. deactivation) measure language function in the brain may underlie the differences seen here.