Beyond role-based surgical domain modeling: Generalizable re-identification in the operating room

Area of Science:

• Computer Science
• Medical Informatics
• Surgical Technology

Background:

• Surgical workflow optimization traditionally focuses on roles, but team familiarity and individual characteristics significantly impact outcomes.
• Existing models often overlook the nuances of individual staff behavior and interaction within the operating room.

Purpose of the Study:

• To develop a novel staff-centric modeling approach for surgical personnel.
• To enable long-term tracking and analysis of individual surgical team members across multiple procedures.
• To improve the accuracy and generalizability of personnel tracking in clinical environments.

Main Methods:

• Characterizing individual team members by distinctive movement patterns and physical characteristics using 3D point clouds.
• Developing a generalizable re-identification framework to encode shape and articulated motion patterns.
• Implementing a novel workflow visualization technique for analyzing team dynamics and space utilization.

Main Results:

• Achieved 86.19% accuracy on realistic clinical data for individual identification.
• Maintained 75.27% accuracy when transferring between different environments, a 12% improvement over existing methods.
• Augmented markerless personnel tracking, improving accuracy by over 50% and addressing occlusions and re-entry issues.

Conclusions:

• The staff-centric model provides novel insights into surgical team dynamics and space utilization.
• This framework advances methods for analyzing surgical workflows and team coordination.
• The approach demonstrates significant improvements in accuracy and generalizability for surgical personnel tracking.