Skip to main content
. Author manuscript; available in PMC: 2020 Jul 1.
Published in final edited form as: Comput Med Imaging Graph. 2019 May 9;75:47–55. doi: 10.1016/j.compmedimag.2019.04.007

Figure 9:

Figure 9:

The top row is the illustration of the differences in the manual labeling protocol of Dataset-3 and the other two datasets. The bottom row is the overlay of automated segmentation of samples images from Dataset-3 with the model trained on Dataset-2 and the ground truth segmentation for these images demonstrating the protocol-based discrepancy lowering the segmentation performance.