Point cloud registration algorithm using liver vascular skeleton feature with computed tomography and ultrasonography...

Area of Science:

• Medical Imaging
• Surgical Navigation
• Computational Geometry

Background:

• Accurate image fusion is crucial for ultrasound-guided liver cancer surgery.
• Conventional methods for fusing ultrasound and computed tomography (CT) images face accuracy challenges.
• Advancements in radiofrequency ablation necessitate improved image registration techniques.

Purpose of the Study:

• To develop a novel algorithm for cross-source point cloud registration between ultrasound and CT images.
• To enhance the accuracy and speed of image fusion for liver cancer interventions.
• To utilize geometric features of vascular point clouds for improved registration.

Main Methods:

• Developed a fusion system for cross-source point cloud registration using ultrasound and CT data.
• Extracted vascular point cloud features including nodes, skeletons, and geometric attributes.
• Implemented a registration algorithm based on these vascular geometric features.

Main Results:

• Achieved a registration error within 1.4 mm in a liver dummy model.
• Significantly reduced target registration error compared to existing methods.
• Demonstrated an average RMSE of 2.23 mm for human liver vascular skeletons.

Conclusions:

• The proposed vascular feature point cloud registration method enables rapid and accurate fusion of ultrasound and CT images.
• This technique is beneficial for real-time puncture surgery in radiofrequency ablation for liver cancer.
• Future work will involve clinical evaluation in patients.