Fast and accurate vision-based stereo reconstruction and motion estimation for image-guided liver surgery

Area of Science:

• Medical imaging
• Computer vision
• Surgical technology

Background:

• Image-guided liver surgery requires precise tumor localization and vasculature mapping for improved oncologic outcomes.
• Existing methods face challenges in handling low-texture regions and motion artifacts common in liver imaging.

Purpose of the Study:

• To present a novel end-to-end solution for fast stereo reconstruction and motion estimation in liver surgery.
• To enhance the accuracy of tumor localization and boundary recovery for improved surgical precision.

Main Methods:

• A computationally efficient coarse-to-fine (CTF) stereo approach with adaptive windows for precise 3D boundary recovery.
• A robust 3D motion estimator to reject spurious data and account for low-texture regions.
• Registration of time-series reconstructions to a key frame for alignment with volumetric computed tomography (CT) scans.

Main Results:

• Demonstrated high accuracy in stereo reconstruction and motion estimation using both phantom and clinical data.
• The adaptive CTF matching approach successfully handled low-texture liver regions.
• The system achieved precise 3D boundary recovery and robust motion estimation.

Conclusions:

• The developed system offers a novel and accurate solution for image-guided liver surgery.
• This approach has the potential to significantly improve surgical precision and patient oncologic outcomes.
• It is the first adaptive CTF matching method for reconstruction and motion estimation that registers time-series data to a key frame for CT registration.