A Technique for Registering ECoG ElectrodesZhongtian Dai1, Vernon L. Towle1, David Brang2, G. Kavya Minama Reddy1, Weili Zheng3

1 The University of Chicago, 2 Northwestern University, 3 The University of Illinois at Chicago

Supported by NIH 5 R01 NS40514, The Brain Research Foundation and The Susman & Asher Foundation

The ProblemThe implantation of electrocorticographic (ECoG), or intracranial EEG (iEEG) electrodes is an invasive process which often introduces non-uniform deformation to a subject’s brain, commonly known as a “brain shift”. As part of the standard clinical procedure, a pre-implant MR image and a post-implant CT image are acquired. But the deformation creates a problem for researchers and clinicians as the locations of electrodes and their recorded activities on the cortical surface are not immediately clear. Thus some processing is required to register the electrodes on the cortical surface. Simply put, if one tries to visualize the electrodes without registering, they would appear buried in the cortex.

Fig. 1. Co-registered pre-implant MR and post-implant CT images shown in gray scale and in orange scale, respectively. The left panel shows the side view of the implanted ECoG electrodes on a patient’s cerebrum. The right panel shows a notable “brain shift” inward of 1-2cm.

Fig. 2. A synthetic electrode, i.e. a disk, and its true and estimated normals shown in black and white respectively. The normal estimated by our principal axis method align well with the true normal (cosine 0.995 with std 0.005). Typically when traveling 1cm along the two normals, the positions will deviate for less than 1mm from each other.

Our SolutionWe developed a novel technique, - we call it the principal axis method -, to register ECoG electrodes by exploiting individual electrodes’ disk-like geometry. The principal axis method treats each electrode as a disk instead of a point, and estimates the normal of individual electrodes by computing the moment of inertia and thus the principal axis of rotation. We then register an electrode to the nearest intersection of its normal and the dura.

To generate the input data, we first co-register the pre-implant MR and the post-implant CT with SPM mutual information image registration. Then we use Freesurfer to extract the dura from the reconstructed pial surface. These are the input to our method.

Fig. 3. Comparison between the principal axis method (green) and the nearest method (red) on real data. Note the difference on electrodes near the posterior ridge.

Fig. 4. Fully registered and color-labeled electrode grids. The upper panel shows the dura and the lower panel shows the pial surface.

ReferencesSPM. SPM software - Statistical Parametric Mapping at <http://www.fil.ion.ucl.ac.uk/spm/software/>FreeSurfer. FreeSurfer at <http://freesurfer.net/>Hunter, J. D. et al. Locating chronically implanted subdural electrodes using surface reconstruction. Clinical Neurophys. 116, 1984–1987 (2005).Tao, J. X. et al. The accuracy and reliability of 3D CT/MRI co-registration in planning epilepsy surgery. Clinical Neurophysiology 120, 748–753 (2009).Hermes, D., Miller, K. J., Noordmans, H. J., Vansteensel, M. J. & Ramsey, N. F. Automated electrocorticographic electrode localization on individually rendered brain surfaces. Journal of Neuroscience Methods 185, 293–298 (2010).