Enhanced CT-based radiomics predicts pathological complete response after neoadjuvant chemotherapy for advanced adenocarcinoma of the esophagogastric junction: a two-center study

Insights into Imaging

Table 2 Efficacy of different models in the training group and the external validation group for predicting pathological complete response (pCR)

Model	Arterial model	Venous model	Clinical model	Arterial–venous combined model	Arterial–clinical combined model	Venous–clinical combined model	Arterial–venous–clinical combined model
Training group
AUC	0.736	0.751	0.753	0.768	0.836	0.818	0.838
95%CI	0.607–0.865	0.614–0.888	0.622–0.884	0.639–0.896	0.728–0.943	0.708–0.927	0.736–0.941
Threshold	0.510	0.856	0.351	0.667	0.332	0.543	0.160
Specificity	0.610	0.829	0.683	0.927	0.805	0.878	0.659
Sensitivity	0.789	0.632	0.789	0.474	0.789	0.632	0.895
Accuracy	0.667	0.767	0.717	0.783	0.800	0.800	0.733
NPV	0.862	0.829	0.875	0.792	0.892	0.837	0.931
PPV	0.484	0.632	0.536	0.750	0.652	0.706	0.548
External validation group
AUC	0.750	0.768	0.848	0.795	0.893	0.884	0.902
95%CI	0.535–0.965	0.489–1	0.710–0.987	0.560–1	0.780–1	0.762–1	0.792–1
Threshold	0.510	0.856	0.351	0.667	0.332	0.543	0.160
Specificity	0.536	0.750	0.857	0.857	0.786	0.893	0.500
Sensitivity	0.750	0.750	0.750	0.500	1	0.500	1
Accuracy	0.562	0.750	0.844	0.812	0.812	0.844	0.562
NPV	0.938	0.955	0.960	0.923	1	0.926	1
PPV	0.188	0.300	0.429	0.333	0.400	0.400	0.222