Interactive Visualization of Machine Learning Model Results Predicting Infection Risk

AutorSchäfer, Steffen; Baumgartl, Tom; Wulff, A.; Kuijper, Arjan; Marschollek, M.; Scheithauer, S.; Landesberger, Tatiana von
ArtConference Paper
AbstraktA high occurrence of infectious diseases in a hospital is a thread for patients and hospital staff. A particular threat are pathogens which have developed resistance to multiple antibiotics as well as the new infections caused by SARS-CoV-2 as part of the worldwide pandemic. Infections occur in outbreaks in a temporally and spatially clustered manner. A promising strategy to reduce new infections is to detect high occurrence of pathogens at an early stage and to trace transmission routes. For clinicians and hygienists (for simplicity ’experts’) it is currently very difficult to monitor the occurrence of infections. Relevant data is only available in tabular format and is neither visually processed nor meaningfully linked. This results in a high amount of time-expensive, manual labor. To help predicting infection risk of a patient, a machine learning model was created and used. The dataset contained over one million test results of patients collected from 2010 to 2014. In order to extract highlevel patterns such as transmission pathways and high pathogen occurence (so-called "clusters") the data needs to be visualized in a compact view.
KonferenzEurographics Conference on Visualization 2022
PublisherThe Eurographics Association
ProjektHiGHmed - Medizininformatik-Konsortium - Beitrag Universitätsklinikum Heidelberg