Calibration-free 3D-2D surface registration for image guided intervention.

Area of Science:

• Medical Imaging
• Computer-Assisted Surgery
• Surgical Robotics

Background:

• Accurate three-dimensional (3D) to two-dimensional (2D) registration is essential for image-guided surgery and augmented reality (AR) applications.
• Conventional methods rely on precise camera calibration, which is difficult to achieve, especially with high-zoom cameras, limiting registration accuracy.

Purpose of the Study:

• To develop a novel 3D-2D surface registration method that eliminates the need for camera calibration.
• To improve the accuracy, robustness, and efficiency of registration for augmented reality in minimally invasive surgery.

Main Methods:

• The proposed method transforms 3D-2D registration into a 2.5D-2.5D process under a small field of view.
• A bias-correction technique is introduced for pseudo depth map estimation, converting multimodal registration to approximate unimodal registration.
• Validation was performed on surgical exoscope and endoscopic datasets.

Main Results:

• The method achieved a mean target registration error of 1.56 mm for rigid phantoms and 1.53 mm for deformable cadaver brains using a high-zoom surgical exoscope.
• For endoscopic data, a 2D fiducial distance error of 1.95 mm was achieved.
• The approach demonstrated robustness across various initial conditions and camera types, including those with unknown parameters.

Conclusions:

• The novel camera-calibration-free method significantly enhances the accuracy and robustness of 3D-2D registration for surgical AR.
• This technique simplifies the registration process, making it more efficient and applicable to challenging scenarios like high-zoom exoscopes and deformable tissues.