Table 2.

Segmentation performance evaluated in mean Dice similarity coefficient (DSC) and standard deviation (SD) with manual segmentation as the ground truth, and DSC prediction performance evaluated in mean absolute error (MAE) and Pearson correlation ($r$). All $r$ had p $<0.0005$.

Segmentation Model	Mean DSC (SD)	MAE	r
U-net 7	0.6688 (0.1715)	0.0312	0.95
U-net 11	0.8091 (0.0738)	0.0368	0.73
U-net 15	0.8301 (0.0678)	0.0380	0.66
U-net 19	0.8264 (0.0547)	0.0384	0.43
U-net 23	0.8309 (0.0556)	0.0382	0.57
U-net 27	0.8313 (0.0578)	0.0399	0.42
Combined Model 1	0.7896 (0.0769)	0.0463	0.60
Combined Model 2	0.8368 (0.0598)	0.0405	0.47
Combined Model 3	0.8371 (0.0565)	0.0382	0.52
Combined Model 4	0.8288 (0.0576)	0.0354	0.73
Combined Model 5	0.8087 (0.0725)	0.0333	0.88
Combined Model 6	0.6883 (0.1779)	0.0335	0.96
QCD	0.8508 (0.0541)	0.0339	0.53
QCD-Lite	0.8503 (0.0562)	0.0344	0.58
Weighted Average QCD	0.8225 (0.0590)	0.0315	0.71