IFAC PapersOnLine 54-4 (2021) 135–140 ScienceDirect ScienceDirect Available online at www.sciencedirect.com 2405-8963 Copyright © 2021 The Authors. This is an open access article under the CC BY-NC-ND license. Peer review under responsibility of International Federation of Automatic Control. 10.1016/j.ifacol.2021.10.023 Copyright © 2021 The Authors. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/) Acute Respiratory Distress Syndrome (ARDS), which was firstly mentioned in literature in the 1960s, is a serious disease with various causes (Ashbaug et al. (1967)). In an epidemiologic study, where Intensive Care Units (ICU) of 50 countries took part, about 10 percent of all pa- tients admitted fulfilled the criteria for an ARDS (Bellani et al. (2016)). A late or even missed recognition of the syndrome might contribute to its high mortality (Bellani et al. (2020)). The most prominent definition of ARDS was presented by the Berlin Taskforce in 2012 and rep- resents the current standard (Rubenfeld et al. (2012)). One criteria of the Berlin Definition defines the distinc- tion or rather exclusion of a heart failure, which usually needs an objective assessment (e.g. echocardiography) by a physician. Echocardiograms or imaging results are not commonly available for every patient and their generation is associated with effort and time. Thus, we propose an algorithmic classification for the retrospective distinction of HF and ARDS. This classification is based on continu- ous vital signs, which are recorded by medical embedded ⋆ This publication of the SMITH consortium was supported by the German Federal Ministry of Education and Research, grant numbers: 01ZZ1803K, 01ZZ1803B and 01ZZ1803M. 1. INTRODUCTION 2. RELATED WORK During the past decades, a variety of computer models have been proposed in literature adressing different fields systems, and laboratory results. Scores, that are calculated with recorded data, are widely adopted in the medical field and can support the physician during the diagnosis process, as p.e. the SOFA-score, which provides an degree of organ dysfunction of a patient (Lopes-Ferreira et al. (2001)). To validate this classification, a database, referred to as ICCA-DB, is provided by the University Hospital Aachen and comprehends 27,256 deidentified patients. As the average time between two datapoints of necessary vital signs in the ICCA-DB is about 170 minutes, a continuous evaluation of the patient status is difficult. Especially the arterial oxygen partial pressure P a O 2 , which is neces- sary to classify the severity of a possible ARDS, is only measured every 248 minutes on average. Therefore, we implemented a model of the cardiovascular and respiratory system to generate this data. As especially mathematical modeling has been widely established (Verma et al. (1981), Cobelli et al. (2014)), the implemented model addresses a systemic approach and relies on different models from literature, which will be described in the following section. Keywords: Medical applications, Classification, Modeling, Diagnostic inference, Computer model, Simulation, ARDS Abstract: Acute Respiratory Distress Syndrome (ARDS) is a common cause for respiratory failure and has a high mortality rate of 30-40% in most studies. The current standard for the diagnosis of ARDS was proposed by the Berlin Definition from 2012. This article proposes an algorithmic classification to distinguish between patients with ARDS and those with heart failure (HF). Currently, the available database is not sufficient in regards to the necessary data quality to evaluate this classification. Therefore an approach of simulating data for patients with ARDS and HF by using a computer model was implemented. The model and classification are evaluated using selected patient data, which is recorded with medical embedded systems in intensive care units, as an input for the simulation. The included scores provide a retrospective assessment of whether or not a patient has developed an ARDS. * Informatik 11 - Embedded Software, RWTH Aachen University, Aachen, Germany (e-mail: fonck@embedded.rwth-aachen.de) ** SMITH Consortium of the German Medical Informatics Initiative, Aachen, Germany *** Department of Intensive Care Medicine, University Hospital RWTH Aachen, Aachen, Germany **** Juelich Supercomputing Centre, Forschungszentrum Juelich, J¨ ulich, Germany Simon Fonck, *,** Sebastian Fritsch, **,***,**** Stefan Kowalewski, * Raimund Hensen, * Andr´ e Stollenwerk *,** Algorithmic distinction of ARDS and Heart Failure in ICU data from medical embedded systems by using a computer model ⋆