Skip to main content
. 2021 Feb 18;10:623818. doi: 10.3389/fonc.2020.623818

Figure 4.

Figure 4

Performance of Deep survival learning (DSL) model. (A) In the independent testing set, we achieved 0.14 of brier score using our DSL model. (B) DSL model made a median absolute error of 1.989 patients and mean absolute error of 2.621 patients during 12,000 day of follow-up time in testing set. (C) DSL model made a median absolute error of 3.851 and mean absolute error of 5.632 in survival curve prediction. Nearly all spots of predicted curve lied within confidence intervals of actual curve and the predicted curve was drew similarly to the actual one.