Laparoscopic Image-Based Critical Action Recognition and Anticipation With Explainable Features

Area of Science:

• Computer Science
• Medical Informatics
• Surgical Technology

Background:

• Surgical workflow analysis is crucial for real-time surgical support systems.
• Accurate recognition and prediction of surgical actions are essential for effective surgical guidance.
• Existing methods lack robustness and interpretability in in-vivo surgical scenarios.

Purpose of the Study:

• To propose a novel framework for fine-grained surgical workflow analysis.
• To improve the robustness and interpretability of critical action recognition.
• To enhance the prediction accuracy of surgical instrument motion tendencies.

Main Methods:

• Developed a framework incorporating operational experience for action recognition.
• Utilized a hierarchical classification structure with an explainable feature space.
• Modeled instrument motion primitives in the polar coordinate system (PCS) for trajectory forecasting.
• Implemented adaptive pattern recognition (APR) to handle laparoscopic surgical variations.

Main Results:

• Achieved exceptional accuracy in critical action recognition.
• Demonstrated real-time performance for surgical awareness tasks.
• Successfully predicted instrument motion tendencies with improved accuracy.

Conclusions:

• The proposed framework significantly enhances surgical awareness through robust action recognition and accurate motion prediction.
• The integration of operational experience and adaptive pattern recognition improves system performance in complex in-vivo settings.
• This approach offers a promising direction for advancing real-time surgical support systems.