Figure 3
From: Overcoming the limitations of patch-based learning to detect cancer in whole slide images
Comparison between Dice and dprim. Each annotation is converted to its convex hull to extract dprim between the compared masks. To avoid crowding, we only show the tumor bed extents for (a), where the longest diagonal difference (shown in red lines) is ∼ 1 mm, whereas the perpendicular diagonal difference is ∼ 2 mm.
