Fig. 4

Prediction performance and calibration of the multiparametric MRI-based nomogram in detecting cervical lymph node metastasis. The nomogram cutoff value of 210 points was used to generate the patients into metastasis risk groups in this study. Prediction performance of the multiparametric MRI-based nomogram in the primary (a) and validation (b) cohorts was measured by the area under the receiver operating characteristic curve. The predictive accuracy to discriminate the patients likelihood of having occult nodal metastasis were 0.952 (95% CI 0.917–0.987, P  < 0.001) in the primary cohort and 0.881 (95% CI 0.779–0.983 (P < 0.001) in the validation cohort. Calibration curve of the MRI-based nomogram for predicting the risk of occult nodal metastasis in the primary (c) and validation (d) cohorts of patients with early-stage SCCLT. Calibration curve indicated there had no apparent departure from perfect fit, with good correlation between the prediction and observation in both cohorts.