Abstract

Cortical language maps, obtained through intraoperative electrical stimulation studies, provide a rich source of information for research on language organization.  Previous studies have shown interesting correlations between the distribution of essential language sites and such behavioral indicators as verbal IQ and have provided suggestive evidence for regarding human language cortex as an organization of multiple distributed systems. Noninvasive studies using ECoG, PET, and functional MR lend support to this model; however, there as yet are no studies that integrate these two forms of information. In this paper we describe a method for mapping the stimulation data onto a 3-D MRI-based neuroanatomic model of the individual patient. The mapping is done by comparing an intraoperative photograph of the exposed cortical surface with a computer-based MR visualization of the surface, interactively indicating corresponding stimulation sites, and recording 3-D MR machine coordinates of the indicated sites. Repeatability studies were performed to validate the accuracy of the mapping technique. Six observers—a neurosurgeon, a radiologist, and four computer scientists, independently mapped 218 stimulation sites from 12 patients. The mean distance of a mapping from the mean location of each site was 2.07 mm, with a standard deviation of 1.5 mm, or within 5.07 mm with 95% confidence. Since the surgical sites are accurate within approximately 1 cm, these results show that the visualization-based approach is accurate within the limits of the stimulation maps. When incorporated within the kind of information system envisioned by the Human Brain Project, this anatomically based method will not only provide a key link between noninvasive and invasive approaches to understanding language organization, but will also provide the basis for studying the relationship between language function and anatomical variability.

r 1997 Academic Press