Back

Generating virtual patient data for in-silico clinical trials of medical devices during extracorporeal membrane oxygenationtreatment

Landoll, M.; Huang, Y.; Follegot, F.; Strassmann, S.; Steinseifer, U.; Karagiannidis, C.; Neidlin, M.

2024-09-07 bioinformatics
10.1101/2024.09.03.610823 bioRxiv
Show abstract

This study introduces a virtual patient generation model as online tool through the generation of high-quality synthetic data, addressing challenges like privacy cocerns and limited dataset sizes. Using a Conditional Tabular Generative Adversarial Network (CTGAN), we generated synthetic data from the Electronic Health Records (EHR) of 767 veno-venous extracorporeal membrane oxygenation (ECMO) patients, focusing on 55 critical therapy parameters. Rigorous preprocessing, imputation, and model tuning ensured that the synthetic data closely mirrored real patient records, achieving a 86.6% coverage score and minimal deviations in data correlations. The tool is integrated into a web platform - https://cve-sim.de/ecmo-vpg, allowing researchers to generate and visualize virtual patient cohorts, potentially enhancing ECMO research and reducing the need for extensive clinical trials.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.