Usage of a graph database for the selection of sterile items in the OR

In this work, we present a subsystem of a robotic circulating nurse, that produces recommendations for the next supplied sterile item based on incomplete requests from the sterile OR staff, the current situation, predefined knowledge and experience from previous surgeries. We describe a structure to store and query the underlying information in terms of entities and their relationships of varying strength.

For the implementation, the graph database Neo4j is used as a core component together with its querying language Cypher. We outline a specific structure of nodes and relationships, i.e., a graph. Primarily, it allows to represent entities like surgeons, surgery types and items, as well as their complex interconnectivity. In addition, it enables to match given situations and partial requests in the OR with corresponding subgraphs. The subgraphs provide suitable sterile items and allow to prioritize them according to their utilization frequency.

The graph database was populated with existing data from 854 surgeries describing the intraoperative use of sterile items. A test scenario is evaluated in which a request for "Prolene" is made during a cholecystectomy. The software identifies a specific "Prolene" suture material as the most probable requested sterile item, because of its utilization frequency from over 95%. Other "Prolene" suture materials were used in less than 15% of the cholecystectomies.

We have proposed a graph database for the selection of sterile items in the operating room. The example shows how the partial information from different sources can be easily integrated in a query, leading to an unique result. Eventually, we propose possible enhancements to further improve the quality of the recommendations. In the next step, the recommendations of the software will be evaluated in real time during surgeries.