Feature-based multimodal 3D/2D and 3D/3D registration framework for pedicle screw registration and evaluation.

Area of Science:

• Medical Imaging
• Spine Surgery
• Computer-Aided Surgery

Background:

• Accurate pedicle screw placement is critical for spine surgery safety.
• Multimodal image registration faces challenges like intensity mismatches and tissue non-correspondence.

Purpose of the Study:

• Develop and validate a feature-based 3D to 2D registration framework.
• Integrate pre-operative MR with intra-operative X-ray for pedicle screw trajectory estimation.
• Achieve robust 3D visualization and overcome multimodal registration limitations.

Main Methods:

• Utilized vertebral endplates as consistent features for registration.
• Employed forward projection of segmented MR images and binary image matching.
• Optimized registration using CMA-ES and validated with 3D to 3D registration against CT.

Main Results:

• The framework accurately estimated pedicle screw trajectories with errors below 2.5 mm.
• Demonstrated precise alignment of trajectory positions across different imaging modalities.
• Validated accuracy through 3D to 3D registration of pre-operative MR and post-operative CT.

Conclusions:

• The feature-based registration framework offers fast and accurate 3D trajectory information for spine navigation.
• This method enhances surgical precision and provides reliable evaluation metrics.
• It serves as a valuable tool for clinical diagnosis in pedicle screw placement procedures.