Development and validation of an ultrasound-based deep learning radiomics nomogram for predicting the malignant risk of ovarian tumours

Table 1 The performance of models predicting the malignancy risk of ovarian tumours

	O-RADS	Clinic_Sig	Rad_Sig	DTL_Sig	DLR_Nomogram
Task 1
Training set
AUC	0.960	0.977	0.952	0.915	0.985
2.5%CI	0.948	0.967	0.938	0.893	0.977
97.5%CI	0.971	0.987	0.967	0.938	0.993
Accuracy	0.881	0.925	0.884	0.860	0.948
Precision	0.754	0.895	0.843	0.819	0.925
Recall	0.983	0.892	0.825	0.775	0.929
F1 score	0.854	0.894	0.834	0.797	0.927
Testing set
AUC	0.960	0.916	0.735	0.890	0.928
2.5%CI	0.948	0.864	0.661	0.839	0.885
97.5%CI	0.971	0.967	0.809	0.941	0.971
Accuracy	0.881	0.859	0.671	0.841	0.871
Precision	0.754	0.875	0.530	0.800	0.880
Recall	0.983	0.700	0.583	0.733	0.733
F1 score	0.854	0.778	0.556	0.765	0.800
Task 2
Training set
AUC		0.896	0.877	0.937	0.955
2.5%CI		0.857	0.835	0.911	0.931
97.5%CI		0.936	0.919	0.963	0.980
Accuracy		0.872	0.789	0.879	0.888
Precision		0.889	0.917	0.896	0.900
Recall		0.949	0.792	0.949	0.958
F1 score		0.918	0.850	0.922	0.928
Testing set
AUC		0.829	0.695	0.853	0.869
2.5%CI		0.728	0.546	0.749	0.765
97.5%CI		0.930	0.845	0.956	0.974
Accuracy		0.808	0.714	0.844	0.779
Precision		0.867	0.846	0.926	0.815
Recall		0.881	0.759	0.862	0.914
F1 score		0.874	0.800	0.893	0.862

Abbreviations: O-RADS ovarian-adnexal reporting and data system, CI confidence interval, Clinic_Sig clinical signature, Rad_Sig radiomics signature, DTL_Sig deep transfer learning radiomic signature, DLR_Nomogram deep learning radiomic nomogram, AUC area under the curve of ROC

ISSN: 1475-925X