Objective:

To develop a robust methodology for tracking soft-tissue regions during surgery, particularly addressing long-horizon occlusions, which are critical for maintaining visibility and accuracy in surgical procedures.

Key Findings:

• The proposed method effectively tracks soft-tissue regions even under long-horizon occlusions, achieving a tracking accuracy of X% as validated on the SurgT benchmark.
• Contrastive learning enhances the robustness of template matching against severe tissue deformations, improving re-identification rates by Y%.
• Validation on the public SurgT benchmark and a synthetic dataset demonstrates the method's effectiveness, with results indicating Z% improvement over existing techniques.

Interpretation:

The integration of dense optical flow and contrastive learning provides a significant advancement in tracking capabilities in surgical environments, particularly for occluded tissues, thereby enhancing surgical precision and safety.

Limitations:

• The approach may still struggle with extremely low-texture tissue surfaces, such as those found in certain abdominal surgeries.
• Performance may vary based on the complexity of the surgical scene, particularly in highly dynamic environments.

Conclusion:

The methodology presents a promising solution for continuous tracking of soft tissues during surgery, addressing challenges posed by occlusions.