Fig. 6: Examples of the segmentation results for the proposed architecture Connected-ResUNets with and without adding the synthetic data.
From: Connected-UNets: a deep learning architecture for breast mass segmentation
Subplot on the top shows two samples of mammograms from the CBIS-DDSM dataset. Subplot on the bottom shows two samples of mammograms from the INbreast dataset. Each rows indicate original mammogram, its corresponding synthetic mammogram, ground-truth (binary) image, segmentation output image of the Connected-ResUNets model trained without adding synthetic data, and segmentation output image of the Connected-ResUNets model trained with adding synthetic data.
