Fig. 3.

Performance of model 3 on set A and set B. Boxplots displaying the Dice similarity coefficient (DSC) (a), the 95th percentile of the Hausdorff distance (HD95; log applied to improve visualization) (b), and the relative volume difference for whole prostate (RVD (WP)) (c) comparing agreement between the masks of the PROSTATEx dataset generated by model 3 with set A and set B, separately. The median DSC and HD95 for whole prostate (WP), peripheral zone (PZ), and transition zone (TZ) were 0.922, 0.755, and 0.878 and 6.34, 10.16, and 8.86 mm, respectively, compared with set A. The medians DSC and HD95 for WP, PZ, and TZ were 0.908, 0.726, and 0.879 and 6.40, 9.80, and 7.78 mm, respectively, compared with Set B. The median RVD (WP) were 8.52% and − 3.19% for model 1 and model 2, respectively. ns: p ≥ 0.05, *** p < 0.001, **** p < 0.0001