Fig. 3

Evaluation of predictive performances for the integrated clinical-deep learning nomogram in classification of lipoma and WDLPS. A Nomogram model combining significant clinical variables, age at diagnosis and serum LDH level, and the deep learning signature. The deep learning signature was generated from the multimodality deep learning-based ResNet50 model with the largest AUC value among all models during external validation. B ROC curves for the predictive performance of the integrated clinical-deep learning nomogram in the training and validation cohorts, respectively. C Precision-recall plots for the predictive performance of the integrated clinical-deep learning nomogram in the training and validation cohorts, respectively. D Curves of the calibration analysis for the integrated clinical-deep learning nomogram in the training and validation cohorts, respectively. E The decision curve analysis for the integrated clinical-deep learning nomogram. WDLPS, Well-differentiated liposarcoma; AUC, Area under the receiver operating characteristic curve; ROC, Receiver operating characteristic