Real-time estimation of FLE statistics for 3-D tracking with point-based registration

Area of Science:

• Medical Imaging
• Surgical Navigation
• Biomedical Engineering

Background:

• Target registration error (TRE) is a key metric in point-based registration for image-guided surgery (IGS).
• Current methods often estimate fiducial localizer error (FLE) a priori, lacking real-time adaptability.
• Robust TRE estimation is crucial for improving accuracy in IGS applications.

Purpose of the Study:

• To develop a real-time method for estimating FLE statistics.
• To enhance the robustness and accuracy of TRE estimation in optical tracking for IGS.
• To improve patient-to-image registration and surgical strategy through real-time TRE feedback.

Main Methods:

• Estimated FLE statistics from real-time fiducial registration error (FRE) statistics.
• Solved a linear system of equations (Ax=b) for FLE covariance parameters.
• Utilized a pre-computed inverse of the tool geometry matrix (A) to minimize real-time computation.

Main Results:

• Monte Carlo simulations showed root mean square FLE within 70-90 micrometers with good FRE estimates.
• Demonstrated a method for robust real-time TRE estimation for optically tracked tools.
• Validated the potential for enhanced patient-to-image registration and surgical guidance.

Conclusions:

• Real-time FLE estimation significantly improves TRE robustness in IGS.
• The proposed method offers a computationally efficient approach for dynamic TRE assessment.
• Accurate TRE feedback empowers surgeons to optimize procedures and enhance overall system accuracy.