Recursive grid partitioning on a cortical surface model: an optimized technique for the localization of implanted...

Area of Science:

• Neurosurgery
• Medical Imaging
• Computational Anatomy

Background:

• Precise localization of subdural electrodes (SDEs) is critical for interpreting intracranial electrocorticography (iEEG) data.
• Brain deformation due to fluid accumulation can compromise electrode localization accuracy on postoperative CT scans.
• Existing methods using pre- and post-implantation scans often yield inaccurate electrode localization.

Purpose of the Study:

• To introduce and evaluate novel 3D mesh modeling techniques for accurate SDE localization.
• To compare the accuracy and reliability of new methods against existing localization techniques.

Main Methods:

• Development of 3D mesh models depicting the pial surface and smoothed pial envelope.
• Manual electrode localization via surgical photographs (highly accurate but time-intensive).
• Automated localization using recursive grid partitioning combined with manually localized electrodes.
• Evaluation of existing methods by applying them to the study's dataset.
• Utilizing automatic parcellation for anatomical electrode labeling and generating inflated cortical surface models.

Main Results:

• Recursive grid partitioning demonstrated the least error (2.0-mm mean error, 6.4-mm maximum error) compared to prior methods (8.2–11.7 mm max, 2.9–4.1 mm mean).
• Methods using both CT and MRI showed lower error than CT-alone methods.
• Significant reduction in localization error (p < 10(-18)) with the novel recursive partitioning method.

Conclusions:

• Novel 3D mesh modeling techniques significantly improve SDE localization accuracy for epilepsy surgery.
• Automated labeling aids in surgical planning and corroborates stimulation mapping results.
• The pial mesh model visualizes unsampled cortical areas, potentially relevant for seizure onset and function.