Efficient stereo image geometrical reconstruction at arbitrary camera settings from a single calibration.

Area of Science:

• Medical Imaging
• Computer Vision
• Surgical Technology

Background:

• Accurate 3D surface reconstruction from stereo images is crucial for surgical navigation.
• Maintaining camera calibration under dynamic operating microscope settings (zoom, focus) presents a significant challenge.

Purpose of the Study:

• To develop an efficient method for sustaining quantitative image-to-physical space mapping for stereo cameras during surgical procedures.
• To eliminate the need for repeated camera recalibration when image acquisition settings change.

Main Methods:

• Images acquired at arbitrary settings (S) are warped to a reference setting (S(0)) using optical flow-derived deformation fields.
• A simple phantom is imaged to obtain deformation fields for various settings.
• 3D surface reconstruction is performed using a single calibration at the reference setting.

Main Results:

• The method achieved an average accuracy of 1.05 mm (along-axis) and 0.59 mm (perpendicular) for phantom images.
• Clinical case results showed average accuracies of 1.26 mm (along-axis) and 0.71 mm (perpendicular) across 47 arbitrary settings.
• The technique demonstrated robustness across diverse surgical scenarios.

Conclusions:

• The developed method efficiently maintains accurate 3D surface reconstruction from stereo microscope images without recalibration.
• This technique has significant potential for improving image-guided surgery and may be applicable to other video-based imaging systems like endoscopes.