Validation study demonstrates accurate entry point localisation in dynamic tracking models, suggesting a reliable low-cost alternative for surgical navigation.
Key Points
To develop a robust, low-cost framework for entry point and pose localization in robot-assisted percutaneous interventions using inexpensive concentric fiducial patches and robust surface fitting.
Designed a vision-based tracking framework utilizing inexpensive concentric fiducial patches.
Implemented an anisotropic iteratively reweighted least squares (IRLS) algorithm combined with a Tukey biweight M-estimator and deterministic soft-weighting to suppress gross outliers and sensor noise.
Achieved a mean absolute error (MAE) of 1.47 mm during dynamic respiratory tracking.
Maintained stable marker detection under challenging conditions including 60% occlusion and low-light environments.