Fig. 5.

ROC curve and DCA curve for estimating the objective response to TACE therapy. a In the NFH, ZHHAJU, and SYUCC cohorts, the deep learning model had an AUC of 0.95, 0.96, and 0.97 for predicting therapy response via patches, respectively. b Based on the predictive probability, the model presented an AUC of 0.95, 0.96, and 0.97 for predicting therapy response in all patients from NFH, ZHHAJU, and SYUCC cohorts, respectively. In the ZHHAJU (c) and SYUCC (d) cohorts, the DCA indicated that when the threshold probability was above 2% and 4%, THE use of the deep learning model for predicting TACE response would gain more benefit than the “treat-all” patients or “treat-none” schemes