Hierarchical particle optimization for cortical shape correspondence in temporal lobe resection.

Area of Science:

• Neurosurgery
• Medical Imaging
• Computational Anatomy

Background:

• Anterior temporal lobe resection effectively treats temporal lobe epilepsy.
• Post-surgical anatomical changes impact follow-up treatment.
• Accurate pre- and post-surgical surface correspondence is crucial but challenging due to surgical alterations.

Purpose of the Study:

• To develop a novel particle-based method for dense, one-to-one shape correspondence between pre- and post-surgical brain surfaces.
• To address challenges in cortical surface registration caused by non-rigid anatomical changes after temporal lobe resection.

Main Methods:

• A novel particle method employing explicit particle adjacency for smooth correspondence.
• Hierarchical optimization of particles to prevent local optima and improve accuracy.
• Evaluation on simulated and actual anterior temporal lobe resection cases using T1-MRI.

Main Results:

• Improved accuracy in simulated data (Hausdorff distance 4.29 mm, Dice 0.841) compared to existing methods.
• Enhanced shape correspondence in actual patient data (parcellation-off ratio 0.061) and reliable cortical thickness change capture.
• Demonstrated superior correspondence smoothness without self-intersection compared to point-wise methods.

Conclusions:

• The proposed particle method provides robust and smooth one-to-one dense shape correspondence for temporal lobe resections.
• Hierarchical optimization enhances efficiency and accuracy.
• The method is reliable for capturing post-surgical cortical thickness changes and outperforms existing approaches.