Pre-processing

Help Guide - Pre-processing

---

Working scenarios

The following table resumes the scenarios in which you may like to localize electrodes. You can combine all possibilities in iElectrodes.

Images	.
Pre-MRI	Post-MRI
Post-CT	Post-CT

Space (atlas)	.
Native space	Normalized space
Freesurfer atlas	Harvard-Oxford atlas

Electrode Type	.	.
Grid (ECoG)	Depth (SEEG)	Grid + Depth (ECoG+SEEG)

Objective	.
Localization coordinates / Visualization	Plan/Simulate implantation
Anatomical labeling	Planing vs. real implantation coordinates

---

Image pre-processing

Using pre- or post-implantation MRI
Sometimes you can choose between pre or post-implantation MRI. There are pros and cons for each.
If you use post-implantation images, you don't need to correct for brain-shift deformations for ECoG grids or strips, but the voxels around the electrodes will have some distortion (small black holes), and atlas errors may occur.
If you use pre-implantation MRI, the image will look great, the atlas will probably be OK, but you will need to correct brain-shift deformations.

Using SEEG depth electrodes, it is assumed that brain-shift is negligible and no correction is needed (even when using the pre-implantation MRI). If you are using post-implantation MRI, the brain atlas parcellation will likely fail.

There is no ideal solution, and it mostly depends on what is more relevant for you.
Using pre- or post-implantation images is exactly the same in the toolbox. Just load one image or the other at the beginning of the project.

Convert your images to NifTI
All images should be loaded in NIfTI format. Dicom images can be converted to NIfTI using dcm2nii (fast, simple, and efficient), SPM, Fieldtrip, and other software.

Native space or normalized space
Images can be in native patient space or in a normalized space, and again there are pros and cons for each approach.

Native space resembles the characteristics of each individual and can be relevant if brain abnormalities are to be studied.
Normalized space (for example, MNI152) is useful to compare the results of different individuals in a common space.
As an example, see SPM processing script /extras/normalize_Images_MNI.m .

One approach to get the best of both worlds is to process the images in native space and then normalize the output. Then, run the localization procedure in the normalized space. In general, the normalization parameters can be easily applied to images and surfaces obtained in the native space.
Briefly, you need to normalize the MRI T1 image to MNI152 space, and in order to avoid erroneous transformations, you need to use a reliable brain mask. This can be done using riboon.nii after processing native images with Freesufer (see [Some Freesurfer notes]). Other approaches are possible, like manual masking.

FreeSurfer pial surfaces and cortical parcellations can be easily loaded in iElectrodes.

Coregistration
CT and MRI images must be coregistered before loading them in iElectrodes. This can be done by many software or toolboxes like SPM or FSL. Images will be resliced in iElectrodes to have the same dimensions (previous to v1.004, images must be resliced before loading). See the provided script /extras/corregister_Images.m to process coregister images with SPM.

More details
More detailed information on the pre-processing steps is available in our open-access paper Blenkmann et al., 2017

---

Back to [Home]
Previous: [Installation] - Next: [Load images]