An automatic markerless registration method for neurosurgical robotics based on an optical camera.

Area of Science:

• Neurosurgical robotics
• Medical imaging
• Computer-assisted surgery

Background:

• Current markerless registration methods for neurosurgical robotics rely on facial surfaces, requiring manual interaction and limiting patient positioning.
• These limitations hinder flexibility and efficiency in robotic-assisted procedures.

Purpose of the Study:

• To develop an automatic, markerless registration technique for neurosurgical robotics.
• To eliminate patient position constraints and manual interaction in the registration process.

Main Methods:

• Utilizing an optical camera on the robot end effector to capture multi-view images of the patient's head.
• Reconstructing head surface point clouds using multi-view stereo vision.
• Employing a manually drawn mark for automatic coarse registration and surface registration for fine alignment, independent of facial landmarks.

Main Results:

• The head surface was acquired with good repeatability accuracy.
• The average target registration error in a head phantom was [Formula: see text].
• The mean surface registration error was [Formula: see text].

Conclusions:

• The proposed method achieves automatic markerless registration across various patient positions.
• Registration accuracy within the head is guaranteed, offering a novel approach for image-to-robot space spatial relationship establishment.