Novel Radiomics Features for Automated Detection of Cardiac Abnormality in Patients with Pacemaker

Table 3

Performance of four classifiers under the best radiomics feature combination of R-wave ultrasound images in the training and test datasets.

Model	Sensitivity (%)	Specificity (%)	Accuracy (%)	AUC
Decision tree
Training	92.00	94.74	93.18	0.912
Test	92.86	87.50	90.91	0.902
SVM
Training	100	92.11	96.59	0.992
Test	100	87.50	95.45	0.929
Random forest
Training	100	94.74	97.73	0.988
Test	100	87.50	95.45	0.964
AdaBoost
Training	94.00	97.37	95.45	0.901
Test	92.86	87.50	90.91	0.909