Towards Autonomous Instrument Tray Assembly for Sterile Processing Applications

TL;DR

This paper tackles automating the sorting and structured packing of surgical instruments into sterile trays in SPD, addressing delays, contamination risks, and instrument damage from manual handling. It introduces an end-to-end robotic pipeline with a calibrated top-down vision system, a 6-DOF Stäubli TX2-60L arm, and a dual electromagnetic gripper, plus a hybrid perception stack using YOLO12n, SAM, and a cascaded Part-Aware ResNet-18 for detection, segmentation, and fine-grained classification. The authors built a custom dataset of $6975$ images over $31$ instrument classes ($80\%$ training) and evaluate a deterministic $O(n)$ packing algorithm with modular dividers, reporting high perception accuracy and statistically significant reductions in tool-to-tool collisions compared with human-packed trays, aided by precise calibration with mean reprojection error $<0.3$ pixels. This work demonstrates a scalable, safety-enhancing step toward automated SPD workflows and outlines concrete directions for future improvements, including multi-orientation packing and non-magnetic instrument handling.

Abstract

The Sterile Processing and Distribution (SPD) department is responsible for cleaning, disinfecting, inspecting, and assembling surgical instruments between surgeries. Manual inspection and preparation of instrument trays is a time-consuming, error-prone task, often prone to contamination and instrument breakage. In this work, we present a fully automated robotic system that sorts and structurally packs surgical instruments into sterile trays, focusing on automation of the SPD assembly stage. A custom dataset comprising 31 surgical instruments and 6,975 annotated images was collected to train a hybrid perception pipeline using YOLO12 for detection and a cascaded ResNet-based model for fine-grained classification. The system integrates a calibrated vision module, a 6-DOF Staubli TX2-60L robotic arm with a custom dual electromagnetic gripper, and a rule-based packing algorithm that reduces instrument collisions during transport. The packing framework uses 3D printed dividers and holders to physically isolate instruments, reducing collision and friction during transport. Experimental evaluations show high perception accuracy and statistically significant reduction in tool-to-tool collisions compared to human-assembled trays. This work serves as the scalable first step toward automating SPD workflows, improving safety, and consistency of surgical preparation while reducing SPD processing times.

Figures (9)

• Figure 1: a) An SPD technician inspecting and assembling a tray. b) A typical freshly opened tray in the Operating Room
• Figure 2: The robotic system working with a human, showing what the robot sees and the top view of the tray during the assembly process
• Figure 3: List of instruments in the dataset by the instrument group
• Figure 4: Training and inference pipeline
• Figure 5: Heat map at layer 4 of ResNet-18. a) failure b) Success