MR radiomics predicts pathological complete response of esophageal squamous cell carcinoma after neoadjuvant chemoradiotherapy: a multicenter study

Table 2 Predictive performance of final models

	AUC	Accuracy (%)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
Clinical model
Training set	0.592 (0.472–0.716)	58.5 (48/82) [53.7–70.7]	74.4 (29/39) [12.8–94.9]	44.1 (19/43) [20.9–100]	54.7 (29/53) [51.0-100]	65.5 (19/29) [55.3–85.7]
Testing set	0.584 (0.441–0.714)	61.6 (45/73) [43.8–75.3]	58.3 (14/24) [16.7–100]	63.3 (31/49) [20.4–98.0]	43.8 (14/32) [36.1–87.5]	75.6 (31/41) [70.3–100]
Radiomics model
Training set	0.968 (0.933–0.992)	92.7 (76/82) [87.8–97.6]	87.2 (34/39) [76.9–97.4]	97.7 (42/43) [93.0-100.0]	97.1 (34/35) [91.4–100.0]	89.4 (42/47) [82.4–97.6]
Testing set	0.885 (0.800-0.958)	82.2 (60/73) [72.6–90.4]	75.0 (18/24) [58.3–91.7]	85.7 (42/49) [75.5–96.0]	72.0 (18/25) [58.6–89.5]	87.5 (42/48) [80.0-95.5]

Data in parentheses are numerators and denominators, with 95% CIs in brackets. AUC, area under the curve; PPV, positive predictive value; NPV, negative predictive value

ISSN: 1470-7330