Viscoelastic testing is a highly recommended cornerstone of modern coagulation medicine, reducing transfusion needs. A disadvantage of viscoelastic tests is the impossibility of making a definitive statement about the platelet count. Therefore, the aim of this retrospective observational study is, on the one hand, to predict the platelet count based on standard ROTEM parameters with the help of several machine learning methods and, on the other hand, to detect a low platelet count ( \<100000 ml-1 and \< 50000 ml-1).
Study Type
OBSERVATIONAL
Enrollment
2,500
Universitätsklinik für Anästhesie und Intensivmedizin
Linz, Austria
Predicition of platelet conentration from ROTEM measurements using machine learning
Several machine learning techniques for the prediction of the platelet concentration from ROTEM parameters (regression approach), namely linear regression, Random Forest, neural network, gradient boosting machine (GBM) and adaptive boosting (ADA) will be assessed. Describing the quality of these prediction models, the mean square error (MSE), the root of the mean of the square of errors(RMSE), the mean absolute error (MAE), and the root mean squared logarithmic error (RMSLE), and the coefficient of determination (R2) will be used.
Time frame: Obtained ROTEM analyses are the baseline at all four centres and patients will be included if platelets were determined concomitantly within three hours on the same day.
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.