CT-based delta-radiomics for predicting pathological response to neoadjuvant immunochemotherapy in esophageal squamous cell carcinoma: a multicenter study

Table 2 Confounder matrix for the four sets in the four models

Predicted results	Actual results		AUC(95% CIs)	Accuracy (%)	Specificity (%)	Sensitivity (%)
	pCR	Non-pCR
Clinical model
Training data set			0.758[0.658–0.858]	71.0	81.8	44.4
pCR	12	12
Non-pCR	15	54
Testing data set			0.615[0.427–0.804]	68.1	76.5	46.1
pCR	6	8
Non-pCR	7	26
Pretreatment radiomics model
Training data set			0.787[0.675-0.900]	75.2	77.3	70.4
pCR	19	15
Non-pCR	8	51
Testing data set			0.621 [0.436–0.806]	57.4	52.9	69.2
pCR	9	16
Non-pCR	4	18
Delta-radiomics model
Training data set			0.827[0.730–0.925]	74.2	68.2	88.9
pCR	24	21
Non-pCR	3	45
Testing data set			0.790[0.646–0.933]	72.3	70.6	76.9
pCR	10	10
Non-pCR	3	24
Mixed model
Training data set			0.847[0.766–0.928]	71.0	68.2	77.8
pCR	21	21
Non-pCR	6	45
Testing data set			0.719[0.567–0.872]	61.7	52.9	84.6
pCR	11	16
Non-pCR	2	18

pCR, pathological complete response; AUC, area under the curve; CI, confidence interval

ISSN: 1471-2342