% IMPORTANT: The following is UTF-8 encoded. This means that in the presence % of non-ASCII characters, it will not work with BibTeX 0.99 or older. % Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or % “biber”. @ARTICLE{Sharafutdinov:1005293, author = {Sharafutdinov, Konstantin and Fritsch, Sebastian Johannes and Iravani, Mina and Ghalati, Pejman Farhadi and Saffaran, Sina and Bates, Declan G. and Hardman, Jonathan G. and Polzin, Richard and Mayer, Hannah and Marx, Gernot and Bickenbach, Johannes and Schuppert, Andreas}, title = {{C}omputational simulation of virtual patients reduces dataset bias and improves machine learning-based detection of {ARDS} from noisy heterogeneous {ICU} datasets}, journal = {IEEE open journal of engineering in medicine and biology}, volume = {5}, issn = {2644-1276}, address = {New York, NY}, publisher = {IEEE}, reportid = {FZJ-2023-01408}, pages = {611 - 620}, year = {2023}, abstract = {Goal: Machine learning (ML) technologies that leverage large-scale patient data are promising tools predictingdisease evolution in individual patients. However, the limited generalizability of ML models developed on single-center datasets,and their unproven performance in real-world settings, remain significant constraints to their widespread adoption in clinicalpractice. One approach to tackle this issue is to base learning on large multi-center datasets. However, such heterogeneous datasetscan introduce further biases driven by data origin, as data structures and patient cohorts may differ between hospitals. Methods: Inthis paper, we demonstrate how mechanistic virtual patient (VP) modeling can be used to capture specific features of patients’states and dynamics, while reducing biases introduced by heterogeneous datasets. We show how VP modeling can be used for dataaugmentation through identification of individualized model parameters approximating disease states of patients with suspectedacute respiratory distress syndrome (ARDS) from observational data of mixed origin. We compare the results of an unsupervisedlearning method (clustering) in two cases: where the learning is based on original patient data and on data derived in the matchingprocedure of the VP model to real patient data. Results: More robust cluster configurations were observed in clustering using themodel-derived data. VP model-based clustering also reduced biases introduced by the inclusion of data from different hospitalsand was able to discover an additional cluster with significant ARDS enrichment. Conclusions: Our results indicate thatmechanistic VP modeling can be used to significantly reduce biases introduced by learning from heterogeneous datasets and toallow improved discovery of patient cohorts driven exclusively by medical conditions.}, cin = {JSC}, ddc = {570}, cid = {I:(DE-Juel1)JSC-20090406}, pnm = {5112 - Cross-Domain Algorithms, Tools, Methods Labs (ATMLs) and Research Groups (POF4-511) / SMITH - Medizininformatik-Konsortium - Beitrag Forschungszentrum Jülich (01ZZ1803M)}, pid = {G:(DE-HGF)POF4-5112 / G:(BMBF)01ZZ1803M}, typ = {PUB:(DE-HGF)16}, pubmed = {39184970}, UT = {WOS:001294340500001}, doi = {10.1109/OJEMB.2023.3243190}, url = {https://juser.fz-juelich.de/record/1005293}, }