Data-Driven Deformable 3D-2D Registration for Guiding Neuroelectrode Placement in Deep Brain Stimulation.

Area of Science:

• Neurosurgery
• Medical Imaging
• Computational Anatomy

Background:

• Deep brain stimulation (DBS) is crucial for treating movement disorders.
• Accurate electrode placement is vital for DBS efficacy and safety.
• Variability in electrode positioning can lead to suboptimal outcomes.

Purpose of the Study:

• To introduce a deformable 3D-2D registration method for precise 3D neuroelectrode guidance.
• To enhance the accuracy of electrode placement in deep brain stimulation procedures.

Main Methods:

• A model-based, deformable algorithm for 3D-2D image registration was developed.
• A parametric 3D model using B-spline curves captured lead design variations.
• Iterative optimization of 16 degrees-of-freedom maximized image similarity between radiographs and model projections.

Main Results:

• Achieved an accuracy of (0.2 ± 0.2) mm in 3D electrode localization.
• Demonstrated robustness to imaging parameters, with accurate localization at ≥20° view separation.
• Enabled accurate localization using 1/10th the dose of standard fluoroscopy.

Conclusions:

• The method allows precise neuroelectrode guidance from two low-dose radiographic images.
• The approach accounts for potential anatomical deformations at the target site.
• Future work includes runtime optimization and clinical data evaluation.