Onsite assessment of airborne viral transmission risks in hospital: A computer vision-based approach considering multi-space occupant behavior

Abstract

The spread of respiratory infectious viruses in hospitals is a significant concern, as it can lead to severe nosocomial infections. However, onsite infection risk assessment in hospitals presents significant challenges due to the complexity of these environments, characterized by multiple spaces and occupants, as well as the fine-grained behaviors of individuals. To address this, the present study proposes a computer vision-based occupant behavior tracking method for onsite assessment of spatio-temporal airborne viral transmission risks in multi-space hospital environments. First, four key indicators were identified to quantify fine-grained occupant behavior, with particular focus on the high occupant density, multi-space nature of hospital settings. Then, video data on occupant behavior were then collected through the hospital's surveillance system. A multi-object tracking model was developed to extract the relevant behavioral indicators. Next, an improved Wells-Riley model that integrates fine-grained occupant behavior and environmental factors is proposed to calculate the risk of individual infections across multiple spaces in the hospital and to assess the impact of occupant behavior on mitigating cross-infection. Finally, the proposed method was successfully applied to two routes—indoor and outdoor—in the outpatient department of a tertiary hospital in Wuhan, China, to analyze the effects of fine-grained occupant behavior and environmental variables on infection risk. This study enables onsite assessment of spatio-temporal airborne viral transmission risks in multiple spaces within the real hospital scenario, which providing valuable insights for controlling the transmission of respiratory viruses and tracing infection sources.