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

Table 6

Robust performance of the system on the BTCV dataset with 30 images, where “sup-pixel” refers to superpixel reconstruction and “” indicates without.

Labeled dataset	Unlabeled dataset	Image process options	Performance by mean DSC (%)		value
			Test adaptation	Test with adaptation
NIH-TCIA	BTCV	Raw images	60.14	71.19	7.91e − 3
		Rotation	63.04	74.66	1.82e − 4
		Sup-pixel	62.35	72.63	9.54e − 4
		Rotation + sup-pixel	63.92	74.97	6.34e − 6

The best performance values are presented in bold.