AI to prevent postoperative infections

Postoperative surgical site infections (POSI) are a major issue in Norway and worldwide, resulting in significant societal costs due to extended hospital stays, increased need for reoperations, and reduced participation in the workforce. Each year, more than 800,000 people are affected by postoperative wound infections in Europe, and approximately 1,500 dies from POSI each year in Sweden.

A high prevalence of POSI also leads to increased antibiotic treatment, which in turn raises the risk of antibiotic resistance. According to the World Health Organization (WHO), 50% of the bacteria causing POSI will be antibiotic-resistant in the near future.

Making the Invisible Visible

Factors like airflow and particles, even though invisible, can significantly impact patients' condition after surgery.

This project developed tools to visualize airflow and particle distribution in operating rooms. The effects of different ventilation solutions (laminar ventilation and mixing ventilation) on particle distribution were measured during simulated operations at NTNU Gløshaugen.Furthermore, the impact of healthcare personnel's movements on airflow and particle distribution in the room was measured and visualized.

These measurements formed the basis for the development of an "Extended Reality" (XR) application, which will be used as a training tool for healthcare personnel. The XR application illustrates how healthcare personnel's actions, including equipment positioning and movement patterns, influence the risk of POSI.

Interdisciplinary domain knowledge and ML

Several modern technologies were combined to carry out this project. A machine learning (ML) model was used to capture healthcare personnel's movements in the operating room.

During an operation, the operating room is packed with equipment, making it difficult for a single camera to capture the surgical staff. Therefore, several cameras were used in the project, each employing the machine learning model to detect people in its field of view. The data from each camera is then assembled to form a fused representation of the healthcare personnel's movements in the operating room. This data is then used in dynamic numerical fluid dynamics (CFD) simulations to accurately visualize airflow and particle distribution in the operating room.

Additionally, a ML model was developed to simulate and visualize airflow in real time. This enabled rapid and intuitive visualization of airflow and particle distribution during various operating scenarios with different movement patterns and equipment positions. The collaborative nature of this project highlights Norconsult's interdisciplinary expertise, incorporating ventilation domain, fluid dynamics, XR and AI solutions into a cohesive system.

Healthcare economy

This research has the potential to dramatically improve patient safety and reduce costs associated with treating POSI.

By using advanced technologies such as machine learning and data analysis to develop visualization tools, healthcare personnel can better understand and manage risk factors like airflow and particle distribution in operating rooms. This can lead to more effective measures to prevent POSI, reduce the need for reoperations and antibiotic treatments, and thereby improve patient outcomes and society's healthcare economy.