TABLE VIII:
Results of our SCO-SSL method when using mean squared error (MSE) loss, Kullback-Leibler (KL) divergence loss, and binary cross entropy (BCE) loss as the consistency loss for semi-supervised learning.
| Losses | UCLA dataset | NIH dataset | ||
|---|---|---|---|---|
| DSC[%] | ASD[mm] | DSC[%] | ASD[mm] | |
| BCE | 91.60(2.37) | 1.02(0.34) | 90.12 (3.61) | 1.23 (0.63) |
| KL | 91.73(2.35) | 1.00(0.34) | 90.00(3.53) | 1.25(0.62) |
| MSE | 91.76 (2.35) | 1.00 (0.34) | 90.04(3.45) | 1.24(0.61) |