Cross-Modality Registration using Bone Surface Pointcloud for Robotic Ultrasound-Guided Spine Surgery

Area of Science:

• Medical Imaging
• Robotics in Surgery
• Spine Surgery Navigation

Background:

• Image guidance using preoperative MRI and intraoperative US improves spine surgery outcomes.
• Robotic US systems (RUSS) enable automated 3D US volume acquisition for better registration.
• Cross-modality discrepancy between MRI and US presents a significant challenge for accurate registration.

Purpose of the Study:

• To develop and evaluate a novel pipeline for accurate per-vertebra registration in spine surgery.
• To address the challenges of cross-modality discrepancy in image-guided spine surgery.
• To enhance the applicability of robotic US systems in real-time surgical navigation.

Main Methods:

• Extraction of spine pointclouds from preoperative MRI and intraoperative 3D US data.
• Development of a pipeline for per-vertebra registration using extracted pointclouds.
• Experimental validation of the pipeline's accuracy, robustness, and time efficiency.

Main Results:

• Achieved a registration accuracy of 1.82 mm (residual root mean square error) and 7.02 mm (Chamfer distance).
• Demonstrated superior robustness to suboptimal initial conditions compared to baseline methods.
• Exhibited good time efficiency, suitable for real-time surgical conditions.

Conclusions:

• The proposed pipeline effectively addresses cross-modality discrepancies in image-guided spine surgery.
• The method shows significant potential for improving the accuracy and reliability of RUSS-guided spine procedures.
• This approach offers a promising solution for enhancing surgical outcomes in complex spine interventions.