An End-to-End Data-Adaptive Pancreas Segmentation System with an Image Quality Control Toolbox

Table 3

Baselines segmentation performance on the NIH-TCIA dataset, where “sup-pixel” refers to superpixel reconstruction.

Model	Image options	Mean DSC (%)
Li et al. [19]	Raw images	83.06 ± 5.57
Cai et al. [20]	Raw images	82.40 ± 6.70
Yu et al. [16]	Raw images	84.50 ± 4.97
Ours	Raw images	83.13 ± 6.93
	Rotation	84.99 ± 4.86
	Sup-pixel	84.35 ± 6.10
	Rotation + sup-pixel	85.45 ± 4.51

The best performance value is presented in bold.