It is often not possible to adequately localize brain tissue responsible for seizure onsets with scalp electroencephalography (EEG). As a result, implanted intracranial electrodes-subdural and depth types- are used to acquire precise localization of seizure onset region(s) and mapping of eloquent cortical function. Acquiring a post-implantation MRI is contaminated with electrode related magnetic artifact. To overcome this problem, a post-implantation Computed Tomography (CT) scan can be acquired and registered to a pre-implantation MRI. However, various causes of brain shift resulting from the surgical implantation procedure create a distortion in brain shape and geometry resulting in inaccurate observed electrode positions relative to MRI depicted brain anatomy. This inaccuracy can cause clinically significant errors in the models used for epilepsy localization and function mapping.

We aim to apply the deformation resulting from a craniotomy and electrode implantation procedure to a pre-surgical brain Magnetic Resonance Imaging (MRI) scan to allow more accurate localization of implanted intracranial electrodes in epilepsy surgical evaluation and treatment.

Using B-Spline transformations and an Advanced Mattes Mutual Information metric, a nonrigid registration was performed between CT scans before and after the implantation procedure. This transformation was subsequently applied to the brain MRI in order to correct for the deformation in 5 subjects.

After the application of this nonrigid transformation there was an average increase in accuracy of 12.5% ± 2.3% (p<0.001) in the ability of depth electrodes to localize gray matter. On average, subdural grid electrodes were 3mm ± 3mm (p=0.06) closer to the locations given by an intraoperative photograph in one patient when the transformation was applied as well.

This method can improve electrode localization relative to brain anatomy on MRI. Further studies will involve the comparison of head models and the accuracy of source localization techniques when this method has been applied prior to forward model calculations.