3D-2D Image Registration in Virtual Long-Film Imaging: Application to Spinal Deformity Correction

Area of Science:

• Medical Imaging
• Computer-Aided Surgery
• Spinal Surgery

Background:

• Intraoperative imaging is crucial for spinal deformity correction.
• Current methods often require multiple views, complicating surgical workflow.
• Accurate 3D localization of surgical implants is essential for successful outcomes.

Purpose of the Study:

• To investigate intraoperative 2D virtual long-film (VLF) imaging for 3D guidance in spinal deformity correction.
• To exploit multi-slot-scan geometry for parallax views and accurate 3D colocalization from a single radiograph.

Main Methods:

• Utilized multi-slot collimator apertures to create fan-beams with disparate views (parallax).
• Employed 3D-2D image registration with known component models (pedicle screws, spinal rods) to estimate pose.
• Conducted experiments using cadaver studies emulating O-arm system geometry.

Main Results:

• Demonstrated the feasibility of multi-slot VLF imaging.
• Quantified the geometric accuracy of 3D-2D registration using VLF acquisitions.
• Achieved a mean target registration error of (2.0±0.7) mm for pedicle screw registration from a single VLF, comparable to multi-view methods.

Conclusions:

• Single-view 3D-2D registration using VLF imaging is a promising solution for image guidance in spinal deformity correction.
• This method simplifies workflow by eliminating the need for multiple-view registration.
• Potential applications extend beyond spine surgery, including long-bone fracture reduction.