Artificial intelligence-based waste detection in robotic surgery: a pilot validation study of computer vision...

Area of Science:

• Surgical Technology
• Computer Vision
• Healthcare Economics

Background:

• Operating room waste, particularly from single-use supplies in robotic surgery, is a significant overlooked cost.
• Current waste tracking methods rely on manual observation, which is inefficient, inconsistent, and lacks scalability.

Purpose of the Study:

• To develop and validate a computer vision-based system (AIMs) for automatic detection and quantification of robotic stapler load waste.
• To assess the accuracy and workflow integration of the AIMs system in a real-world surgical setting.

Main Methods:

• A prospective validation study involving 18 robotic procedures at a high-volume center.
• Utilized a convolutional neural network for real-time object detection from continuous back table video recordings.
• Compared system-generated logs against concurrent manual observations, evaluating sensitivity and precision at critical timepoints.

Main Results:

• The AIMs system achieved perfect (100%) sensitivity and precision for initial load detection and final waste quantification.
• Mid-procedure tracking demonstrated high sensitivity (87.7%–97.1%) and perfect (100%) precision for addition and removal events.
• All errors were false negatives; no false positives were recorded. System operation did not disrupt workflow or require extra staff.

Conclusions:

• The computer vision-based AIMs system provides accurate, scalable, and objective data for managing surgical supply waste.
• This technology supports preference card optimization and cost reduction initiatives in robotic surgery programs.
• Enhanced cost-effectiveness can facilitate broader patient access to advanced robotic surgical technologies.