Combining Visual Cues with Interactions for 3D-2D Registration in Liver Laparoscopy.

Area of Science:

• Medical Imaging
• Surgical Technology
• Computer-Assisted Surgery

Background:

• Accurate 3D liver model registration is essential for augmented reality (AR) in monocular laparoscopy.
• Challenges include significant preoperative-intraoperative shape differences and limited intraoperative liver visibility.
• Existing methods rely on manual registration, rigid models, or purely automatic visual/biomechanical approaches.

Purpose of the Study:

• To develop and evaluate a novel hybrid approach for 3D liver model registration in laparoscopic surgery.
• To combine machine perception via visual cues with surgeon's spatial understanding through user interaction.
• To improve registration accuracy and repeatability for enhanced surgical guidance.

Main Methods:

• A hybrid registration approach integrating visual cues and user interaction was developed.
• The method leverages machine perception for initial alignment and surgeon input for refinement.
• Registration accuracy and repeatability were assessed using phantom, ex vivo animal, and patient data.

Main Results:

• The proposed hybrid registration method demonstrated superior performance compared to state-of-the-art techniques.
• Significant improvements in both registration accuracy and repeatability were achieved.
• An average registration error below 1 cm, the recommended oncologic margin for laparoscopic hepatectomy, was obtained.

Conclusions:

• The hybrid approach effectively addresses the challenges of 3D liver model registration in AR laparoscopy.
• This method enhances surgical precision by integrating automated perception with expert user knowledge.
• The achieved accuracy supports its clinical utility for oncologic liver resections, potentially improving patient outcomes.