Fig. 4From: Diagnosing capillary leak in critically ill patients: development of an innovative scoring instrument for non-invasive detectionMachine learning based prediction model based on ICU data from day 1. A, B show a decision tree model which was computed from the 7-parameter input variables showing a reliable differentiation using only the two parameters of echogenicity and angiopoietin-2 level. C shows the comparison of diagnostic specifications of different machine-learning models computing the 7-parameter approach (echogenicity, SOFA-Score, angiopoietin-2, syndecan-1, ICAM-1, lactate, IL-6)Back to article page