. Author manuscript; available in PMC: 2021 Jun 1.

Published in final edited form as: IEEE Trans Med Imaging. 2019 Dec 13;39(6):1856–1867. doi: 10.1109/TMI.2019.2959609

TABLE V:

Redesigned skip connections improve both semantic and instance segmentation for the task of nuclei segmentation. We use Mask R-CNN for instance segmentation and U-Net for semantic segmentation in this comparison.

Architecture	Backbone	IoU	Dice	Score
U-Net	resnet101	91.03	75.73	0.244
UNet++	resnet101	92.55	89.74	0.327

Mask R-CNN [12]	resnet101	93.28	87.91	0.401
Mask RCNN++^†	resnet101	95.10	91.36	0.414

^†

Mask R-CNN with UNet++ design in its feature pyramid.