Machine Learning in pulmonary function diagnostic

Due to the current COVID-19 pandemic, people’s state of health is moving more and more into the focus of society.
As COVID-19 is a respiratory disease, patients suffering from chronic obstructive pulmonary disease, colloquially known as smoker’s cough, are classified as at-risk patients. Unfortunately, early detection of the disease in its early stages is only possible to a limited extent even for experts in the context of normal routine preventive examinations with a reliable measurement of the pulmonary function values, as the measurement is highly dependent on the procedure and the patient.

In the context of a project of FHWS in cooperation with the University of Greifswald and Lothar Medtec GmbH as well as Empolis Information Management GmbH, it was investigated to what extent a machine learning model can support the diagnosis of the early stage of COPD. A special requirement here was the interpretability of the model for the treating physicians. Thanks to the cooperation, the students had access to extensive anonymised data from the SHIP study.