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

Table 4

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

Model	Sensitivity (%)	Specificity (%)	Accuracy (%)	AUC
Decision tree
Training	92.00	86.84	89.77	0.883
Test	100	87.50	95.45	0.924
SVM
Training	100	92.11	96.59	0.999
Test	100	87.50	95.45	0.955
Random forest
Training	100	89.47	95.45	0.955
Test	100	87.50	95.45	0.945
AdaBoost
Training	100	89.47	95.45	0.947
Test	92.86	87.50	90.91	0.909