Construction and validation of a predictive model for intracardiac thrombus risk in patients with dilated cardiomyopathy: a retrospective study

Table 2 The performance of different models in predicting the ICT probability in patients with DCM

Models	Datasets	Sensitivity (95%CI)	Specificity (95%CI)	AUC (95%CI)	Accuracy (95%CI)
Logistic Regression	Training set	0.773 (0.685–0.860)	0.803 (0.770–0.837)	0.854 (0.811–0.896)	0.799 (0.768–0.830)
Logistic Regression	Testing set	0.704 (0.531–0.876)	0.827 (0.770–0.883)	0.823 (0.733–0.914)	0.810 (0.756–0.864)
SVM	Training set	0.773 (0.685–0.860)	0.623 (0.582–0.664)	0.769 (0.715–0.824)	0.644 (0.607–0.681)
SVM	Testing set	0.667 (0.489–0.844)	0.653 (0.582–0.724)	0.745 (0.645–0.845)	0.655 (0.589–0.721)
Random Forest	Training set	0.886 (0.820–0.953)	0.807 (0.774–0.840)	0.917 (0.887–0.947)	0.818 (0.788–0.848)
Random Forest	Testing set	0.815 (0.668–0.961)	0.838 (0.783–0.893)	0.880 (0.815–0.945)	0.835 (0.784–0.886)
XGBoost	Training set	0.932 (0.879–0.984)	0.846 (0.815–0.876)	0.947 (0.924–0.969)	0.858 (0.830–0.885)
XGBoost	Testing set	0.815 (0.668–0.961)	0.879 (0.830–0.927)	0.922 (0.865–0.979)	0.870 (0.823–0.917)

Note: ICT, intracardiac thrombus; DCM, dilated cardiomyopathy; SVM, support vector machine; XGBoost, eXtreme Gradient Boosting; AUC, the area under the receiver operating characteristic curve; CI, confidence interval

ISSN: 1471-2261