Normalization of intracranial electrode locations

BLENKMANN ALEJANDRO; PHILLIPS HOLLY; PRINCICH JUAN PABLO; KOCHEN SILVIA

Buenos Aires

Congreso; Latin American Congress on Epilepsy 2014; 2014

International League Against Epilepsy

ObjectiveSpatial localization of intracranial electrodes is needed in order to generate cortical activations maps or applying source localization techniques from EcoG/SEEG recordings. In group analysis comparisons electrode coordinates must be normalized. We developed a processing pipeline based on RMI and CAT images to detect and transform coordinates to a normalized space.MethodsProcessing steps are:1) Coregister MRI and CAT using an affine transformation (SPM8)[1]2) MRI segmentation to obtain a brain mask(FreeSurfer)[2]3) MRI and CAT coregister to MNI-152 space using previous mask and non linear warping deformations (SPM8)[1]4) MRI segmentation to obtain in-skull mask(FSL-BET)[3]5) Electrode coordinates detection by thresholding CAT high intensity voxels, using previous mask and then clustering[4]6) Semi automatic electrode labeling by coordinate projection to expected locations[4]Coregistration error is quantified by measuring the distance from ECoG electrodes to the normalized cortical surfaceResultsSix patients were studied with 104/332 SEEG/ECoG electrodes. RMSE was 4.5mm and standard deviation 2.5mm.ConclusionThe technique showed expected results, even when subdural edema is as big as 21mm (maximum case). The obtained error is admissible if we consider the 4mm diameter of electrodes.In the near future we plan to obtain better results using more robust normalization algorithms.References[1] Ashburner. PhD thesis, UCL,2000[2] Dale, Fischl, Sereno. Neuroimage 9:179-194,1999[3] Smith. HBM, 17(3):143-155,2002[4] Blenkmann et al, CAN,2010