Skip to main content
. Author manuscript; available in PMC: 2023 Jun 1.
Published in final edited form as: IEEE Trans Med Imaging. 2022 Jun 1;41(6):1331–1345. doi: 10.1109/TMI.2021.3139999

TABLE V:

Results of a fully-supervised 3D UNet trained with Shadow-AUG and Shadow-DROP mechanisms using different shadow threshold τs. Bold values indicate the best result. Underlined values suggest results without statistical significance compared with the best result (p>0.05).

Models UCLA dataset NIH dataset
DSC[%] ASD[mm] DSC[%] ASD[mm]
τs = 20/255 92.21 (2.45) 0.94 (0.32) 89.68 (3.74) 1.29(0.65)
τs = 40/255 92.17(2.34) 0.94(0.32) 89.85 (3.30) 1.26 (0.58)
τs = 60/255 92.25 (2.19) 0.93 (0.29) 89.81 (3.86) 1.27 (0.65)
τs = 80/255 92.25 (2.34) 0.93 (0.31) 89.84 (3.90) 1.27 (0.60)