Hepatic Alveolar Echinococcosis: Predictive Biological Activity Based on Radiomics of MRI

Table 3

Diagnostic performance of machine learning-based MRI radiomics classifiers to assess the bioactivity of HAE lesions in the training cohort.

	AUC	Accuracy	Sensitivity	Specificity	95% CI, AUC	Cutoff
LR
0-fold	0.898	0.868	0.861	0.875	0.836-0.942	0.517
1-fold	0.861	0.806	0.903	0.708	0.793-0.913	0.436
2-fold	0.855	0.819	0.819	0.819	0.786-0.908	0.484
3-fold	0.822	0.757	0.889	0.625	0.750-0.881	0.392
4-fold	0.842	0.778	0.708	0.847	0.772-0.898	0.563
Mean		0.806	0.836	0.775	0.750-0.942
MLP
0-fold	0.900	0.875	0.847	0.903	0.839-0.944	0.544
1-fold	1.000	1.000	1.000	1.000	0.975-1.000	0.834
2-fold	0.991	0.958	0.986	0.931	0.958-1.000	0.387
3-fold	0.861	0.778	0.778	0.778	0.793-0.913	0.527
4-fold	0.888	0.819	0.806	0.833	0.825-0.935	0.488
Mean		0.886	0.883	0.889	0.793-1.000
SVM
0-fold	0.898	0.868	0.861	0.875	0.836-0.942	0.517
1-fold	0.861	0.806	0.903	0.708	0.793-0.913	0.436
2-fold	0.855	0.819	0.819	0.819	0.786-0.908	0.484
3-fold	0.822	0.757	0.889	0.625	0.750-0.881	0.392
4-fold	0.842	0.778	0.708	0.847	0.772-0.898	0.563
Mean		0.806	0.836	0.775	0.750-0.942