An Automatic Spatial Registration Method for Image-Guided Neurosurgery System.

Area of Science:

• Neurosurgery
• Medical Imaging
• Computer-Aided Surgery

Background:

• Image-guided neurosurgery systems (IGNS) rely on accurate patient-to-image spatial registration.
• Current methods often involve manual landmark selection, which can be time-consuming and prone to error.

Purpose of the Study:

• To investigate the feasibility of an automatic, marker-free spatial registration method for IGNS.
• To evaluate the performance of the 4-points congruent sets (4PCS) and iterative closest point (ICP) algorithms for this application.

Main Methods:

• A portable scanner acquired patient head point clouds.
• The 4PCS algorithm performed initial, robust registration between patient and image spaces.
• An iterative closest point (ICP) algorithm refined the registration for high accuracy.

Main Results:

• Phantom experiments yielded mean target registration errors of 1.02 ± 0.18 mm (rigid) and 1.27 ± 0.36 mm (elastic).
• Clinical experiments with 3 patients showed mean errors ranging from 1.84 ± 0.19 mm to 1.88 ± 0.19 mm, meeting clinical requirements.
• The proposed method demonstrated superior accuracy and reduced registration time compared to manual registration techniques.

Conclusions:

• The automatic 4PCS and ICP-based spatial registration is feasible for IGNS.
• This method can potentially replace current manual landmark-based registration, enhancing efficiency and accuracy in neurosurgical procedures.