Registration for frameless brain surgery based on stereo imaging.

Area of Science:

• Medical Imaging
• Computer Vision
• Neurosurgery

Background:

• Accurate patient-specific geometric models are crucial for frameless neurosurgery.
• Existing registration methods can be computationally intensive.
• Need for precise facial surface capture for surgical navigation.

Purpose of the Study:

• To implement stereo vision for capturing patient facial geometry for frameless neurosurgery registration.
• To reduce computational requirements for on-site registration.
• To validate the accuracy of the developed system.

Main Methods:

• Stereo vision techniques were employed to capture the 3D geometric model of the patient's face.
• A distance transform was applied to 2D CT/MRI multi-slices for efficient on-site registration.
• A custom phantom was designed to measure the target registration error (TRE).

Main Results:

• The implemented stereo vision system successfully captured facial geometric models.
• On-site registration was achieved with reduced computational load.
• The measured target registration error (TRE) was 2.72 ± 0.735 mm.

Conclusions:

• Stereo vision provides an accurate method for facial geometric model capture in neurosurgery.
• The distance transform enhances registration efficiency for intraoperative use.
• The system demonstrates sufficient accuracy for frameless neurosurgical applications.