Fast two dimensional to three dimensional registration of fluoroscopy and CT-scans using Octrees on segmentation maps

Area of Science:

• Medical Imaging
• Computer-Aided Surgery
• Computational Geometry

Background:

• 3D-to-2D medical image registration is essential for image-guided surgery.
• Digital Reconstructed Radiograph (DRR) generation is computationally intensive, limiting real-time applications.
• Current methods face challenges with large volumetric datasets.

Purpose of the Study:

• To develop a computationally efficient algorithm for generating Digital Reconstructed Radiographs (DRRs).
• To improve the speed of 3D-to-2D medical image registration for virtual surgery applications.
• To enable faster and more accurate image guidance during surgical procedures.

Main Methods:

• Segmentation of volumetric data into regions for compact representation using Octree Data Structures.
• Efficient storage and retrieval of attenuation indexes for DRR projection.
• Maximization of a Mutual Information-based functional for DRR-to-fluoroscopy alignment.

Main Results:

• Demonstrated a computationally efficient DRR generation process.
• Successfully applied the algorithm to virtual surgery scenarios.
• Achieved promising experimental results on real-world medical data.

Conclusions:

• The proposed Octree-based approach significantly reduces DRR generation time.
• This method enhances the feasibility of real-time 3D-to-2D registration in image-guided surgery.
• The algorithm offers a practical solution for improving virtual surgery performance.