A new markerless patient-to-image registration method using a portable 3D scanner.

Area of Science:

• Neurosurgery
• Medical Imaging
• 3D Scanning Technology

Background:

• Patient-to-image registration is crucial for image-guided neurosurgery.
• Conventional methods like skin marker-based registration are costly and time-consuming.
• Existing surface-matching methods using facial scans have limited accuracy, especially for posterior head regions.

Purpose of the Study:

• To introduce a novel surface-matching method for patient-to-image registration using a portable 3D scanner.
• To acquire a complete 3D point cloud of the entire head for improved registration accuracy.
• To overcome the limitations of current registration techniques in neurosurgery.

Main Methods:

• Developed a method to transform scan points from device space to patient space without calibration or tracking.
• Utilized five positioning targets on a reference star for spatial transformation calculation.
• Employed a three-step registration algorithm to align patient space with image space.
• Evaluated the method on rigid and elastic head phantoms and in a clinical patient experiment.

Main Results:

• Achieved mean registration errors of 0.33-0.34 mm (device to patient space) and 1.17-1.34 mm (surface registration).
• Demonstrated high accuracy across all head regions, with similar target registration errors (TREs) in posterior areas.
• Reported a mean TRE of 1.48 ± 0.21 mm on an elastic phantom, outperforming existing facial-only methods (1.98 ± 0.53 mm).
• Clinical experiment yielded a mean TRE of 1.92 ± 0.18 mm, meeting clinical requirements.

Conclusions:

• The proposed portable 3D scanner method offers sufficient registration accuracy, including in the posterior head.
• Acquisition of the entire head's 3D point cloud is feasible and straightforward.
• The method eliminates the need for prior scanner calibration or optical tracking during scanning.