Abstracts

Rationale: Image-guided neurosurgery has proven to be an effective approach to treatment of localizable epilepsy. In recent years, a variety of commercially available Surgical Navigation Systems (SNS) have become more widely used toward this end. These systems allow the surgeon to view co-registered functional, anatomical, and chemical data in real-time relative to the location of surgical tools, and ultimately aid in delineating resection margins. The accuracy of the visualization is limited by brain deformation, registration error, imaging artifact, and post-operative shift. As such, an accurate method for measuring the degree of error in image data is essential to a comprehensive understanding of the intra-operative visualization. Methods: One surgical stage in patients undergoing treatment for medically intractable epilepsy is the implantation of intra-cranial electrodes for EEG measurements on the brain surface (figures 1). Patients were imaged with MRI before surgery and MRI and CT after surgery. Due to the high degree of variability in brain shift, the standardized image space used for all surgical procedures was the pre-operative MRI. Intra-operative landmarks corresponding to specific electrode locations were acquired in pre-operative MR space once the electrode grid was sutured to its final location. These were acquired using the BrainLAB VV Cranial Navigation System and the VVLink research interface. The number of points acquired for each of the four patients varied with the grid size. Electrodes can be easily identified in post-operative images due to the inherent sensitivity of CT to metal. From the CT images, electrode locations were mapped to a coordinate plane in CT space. The intra-operative landmarks in pre-operative MR space were rigidly registered to the post-operative CT containing the electrodes via the post-operative MR (figure 2). Average error was calculated by comparing the coordinates of the registered intra-operative landmarks to the post-operative electrode localizations. This calculation was weighted by the number of points acquired for each patient. Results: A comparison of four patients with an overall assessment of 92 electrode placements yield a weighted average error of 8.0 mm. These errors ranged from 3.3 mm for the most accurate localization to 16.3 mm for the largest error. The overall variability of these error measurements is approximately .5 mm.Conclusions: We have shown a method to measure the accuracy of electrode placement during intra-cranial EEG. Furthermore these results also illustrate the overall error involved in intra-operatively using pre-operative images containing functional information. Because of the variability in brain deformation between patients, the most accurate SNS should take advantage of biomechanical brain-shift models to accurately simulate the location of the brain surface. (Source of Funding: NIH Grant 2R01EB000473-06)