Abstracts

Rationale: The mapping of cortex by the passage of controlled currents - cortical stimulation mapping (CSM) - has been held as the gold standard technique for the localization of eloquent language sites. Despite the widespread use of this technique for several decades, these data are typically represented schematically, referenced to cortical surface landmarks, or to imaginary grids on a representative surface. This has precluded the creation of useful population maps that could be used to identify unusual language organization, to create apriori predictive models to delineate language deficits with focal lesions, or for programmatic comparison with functional imaging data. Methods: 18 patients scheduled for intra-operative (awake craniotomy) language mapping, underwent pre-op hi-res anatomical data acquisition on a 3T scanner. Mesh models of the gray/white matter interface and the pial surface were generated (Dale ’99). A 12 parameter affine transformation was applied to spatially normalize each patient’s anatomical MRI to the T1 weighted MNI single subject template (Collins 1994). Sites of stimulation associated with a behavioral disruption (positive-sites) and no change (negative-sites), were localized on the pial mesh-model using data from an intra-op frameless steretoactic system, intra-op video and in-house software. To depict CSM data, we created a model of the volumes of cortical depolarization from direct current spread in the human brain, fitted by experimental data from surface stimulation of primate cortex (Tehovnik ’96). Indirect or downstream effects of the stimulation, were disregarded. The volume depolarized was constrained by a stimulation mask created for each site of stimulation, that included a finite segment of the cortical ribbon related to the stimulation site. We then used the matrix of transformation generated for image normalization to spatially transform CSM for each individual into standardized space. Using data from positive and negative stimulation results, the point probability of the presence of function was computed for the entire cortical region subjected to electrical stimulation mapping. Data from all patients were averaged to create the population CSM map. Results: A total of 158 positive-sites (Figure 1) and 160 negative-sites were localized using CSM. ELS were found densely clustered and localized exclusively to STG, mid and posterior MTG, supramarginal gyrus, angular gyrus, Broca’s area and the posterior middle frontal gyrus (Figure 2). Three ‘maxima’ were found in this cluster; posterior STG (BA42), mid STG (BA22), and Broca’s Area (BA44). As expected negative stimulation sites were sparsely clustered and did not overlie the maxima seen in the ELS cluster. Conclusions: Spatially probability maps of language, such as this, allow for direct comparison of CSM data with studies derived from functional imaging and lesion analysis. Additionally, the availability of larger numbers of ELS and negative stimulation sites may allow for estimates of risk to language during cortical resections and various brain lesions.