Fig. 3.

The boxplots summarize the comparison between the three variants of the trained networks: sensitivity, Dice coefficient, FP count (visualized using logarithmic scale) and FPR (log scale) for all 89 ZURICH data sets were compared between the nets trained on the routine data (Rou1_Rou2, green), trained on the Philips Ingenia data with pre-trained encoder (Rou1_PhIng, blue) and trained completely on the Philips Ingenia data (PhIng_PhIng, red). See Section 2.4.2 for the description of the variants. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)