Contrast-enhanced CT-based radiomics model for differentiating risk subgroups of thymic epithelial tumors

BMC Medical Imaging

Table 2 The prediction performance of the two RMs

Simplified groups		Subgroups	AUC (95%CI)	Accuracy	Precision	Recall	F1-score
Traditional risk grouping	Training set	LRT (A,AB,B1)	0.795 (0.731–0.859)	0.65	0.73	0.67	0.70
		HRT (B2,B3)	0.851 (0.788–0.900)	0.65	0.63	0.57	0.60
		TC	0.860 (0.797–0.906)	0.65	0.58	0.72	0.64
	Testing set	LRT (A,AB,B1)	0.621 (0.434–0.768)	0.47	0.70	0.58	0.64
		HRT (B2,B3)	0.754 (0.583–0.898)	0.47	0.46	0.50	0.48
		TC	0.500 (0.320–0.683)	0.47	0.27	0.30	0.29
Improved risk grouping	Training set	LRT* (A,AB)	0.855 (0.802–0.906)	0.72	0.73	0.59	0.65
		HRT* (B1,B2,B3)	0.862 (0.780–0.931)	0.72	0.74	0.85	0.79
		TC	0.869 (0.815–0.916)	0.72	0.65	0.56	0.60
	Testing set	LRT* (A,AB)	0.716 (0.554–0.868)	0.62	0.40	0.29	0.33
		HRT* (B1,B2,B3)	0.778 (0.591–0.935)	0.62	0.61	0.82	0.70
		TC	0.879 (0.760–0.983)	0.62	0.83	0.50	0.62

RM, radiomics model; LRT, low-risk thymomas; HRT, high-risk thymomas; TC, thymic carcinoma; AUC, area under the curve

ISSN: 1471-2342