Coregister post-insertion CT images for SEEG analysis

Simeon Wong
Ibrahim Lab @ SickKids

Automatically mark electrode contacts in post-insertion CT imaging, roughly cluster into electrodes, then warp coordinates into standard MNI space.

Installation

• requires FMRIB Software Library (FSL)
• requires X Windows System with OpenGL (eg. direct desktop access or VirtualGL headless)

pip install git+https://github.com/gmilab/seegloc

Usage

seegloc_coreg /path/to/ct.nii.gz /path/to/mri.nii.gz /path/to/subject/coreg_output_dir
seegloc_autolabel /path/to/subject/coreg_output_dir

If this code has been useful in your work, please cite:

Wong SM, Arski ON, Ibrahim GM. An automated algorithm for stereoelectroencephalography electrode localization and labelling. Seizure: European Journal of Epilepsy. 2024;117:293-297. doi:10.1016/j.seizure.2024.04.002

Outputs

Electrode coordinate table

ename	x	y	z	aal	entry
A-01	-43.3236	2.77232	40.8664	Precentral_L	FC3
A-02	-38.8153	2.01561	38.612	Precentral_L	FC3
A-03	-34.0798	1.52495	36.5061	Precentral_L	FC3
A-04	-28.3241	0.982843	34.7693	Frontal_Mid_2_L	FC3
A-05	-19.7692	-1.11625	33.8938	Cingulate_Mid_L	FC3

3D rendering of labelled electrodes

Coregistration QC image (CT overlaid on MRI template)

Troubleshooting

Manual transforms

Automatic electrode segmentation from CT may not always work in some cases where individual electrodes are not easily identified (eg. electrode is so bright there is no space between electrode contacts).
In this case:

1. Using CT viewing software (eg. fsleyes), note the CT scanner-space coordinates in mm of:
   1. The tip (deepest contact) of the electrode trajectory
   2. Any other point along the trajectory
2. Run the seegloc_extrapolate command:

seegloc_extrapolate [spacing in mm] [number of contacts] \
                    [tip_x] [tip_y] [tip_z] \
                    [other_x] [other_y] [other_z] \
                    --coreg /path/to/subject/coreg_output_dir

Algorithms

• scipy label / regionprops for contact identification
• custom code for electrode clustering
• fsl flirt (linear coregistration) of CT <=> Subject MRI
• fsl fnirt (non-linear coregistration) of Subject MRI <=> MNI152 template