Operating rooms are a major cost factor in a hospital's budget. Therefore, there is a need for process optimization related to the operating rooms (OR). However, the collection of key figures for process optimization is often done manually by medical staff. This can be erroneous, inaccurate, time consuming, and incomplete. Automated, data-driven approaches are intended to address these problems and help to get the most precise picture possible of what is happening within the OR. At Heidelberg University Hospital (UKHD), a distributed AI based streaming analytics architecture was set up and integrated into the Medical Data Integration Center (MeDIC). This architecture can process, store, and visualize heterogeneous data from different sources. Data from medical devices and the video stream of the wall mounted cameras of four integrated operating rooms are ingested into our system. Aggregated and analyzed in real-time computed key figures including OR state and utilization numbers are visualized in a dashboard for monitoring and decision support. Because of high data protection hurdles the proposed system, especially the video analytics, was trained and tested with statists and did not run during real procedures. Studies to evaluate and test the system during live surgeries are planned.
Keywords: Artificial Intelligence; Data Management; Interoperability; MeDIC; Operating Room Information Systems.