Toward real-time endoscopically-guided robotic navigation based on a 3D virtual surgical field model

Area of Science:

• Robotics
• Surgical Navigation
• Medical Imaging

Background:

• Accurate guidance of surgical tools is critical in robotically-assisted operations, particularly for procedures like brain tumor margin removal.
• Existing methods face challenges in precisely navigating within complex 3D surgical environments.

Purpose of the Study:

• To develop and evaluate a 3D image-guided surgical navigation technique for robotically-assisted surgery.
• To enable precise guidance of surgical tools using intraoperative video feedback matched to a 3D virtual model.

Main Methods:

• A laser-scanning endoscopic camera attached to a mock surgical tool was used to capture video frames.
• Features from intraoperative video frames were matched to a pre-reconstructed 3D virtual model of the surgical field.
• A constrained bundle adjustment algorithm was implemented to recover camera pose (position and orientation).
• Navigational error was assessed by comparing calculated and measured camera poses.

Main Results:

• The developed algorithm demonstrated near real-time computation efficiency (2.5 seconds per pose estimation in MATLAB).
• Average navigational errors were found to be 3 mm for distance and 2.5 degrees for orientation.
• Identified error sources include 3D model inaccuracy, endoscope parameter inaccuracies, and image distortion.

Conclusions:

• The study demonstrates the feasibility of using a micro-camera for 3D guidance of robotic surgical tools.
• The proposed 3D image-guided navigation system shows potential for improving accuracy in robotically-assisted surgical procedures.
• Further optimization, such as implementation in C++, could enhance computational efficiency for clinical application.