Table 3.

Comparison of dice coefficients of segmentation models.

Model	Normal (n = 102)	Coronal (n = 59)	Sagittal (n = 58)	Lambdoid (n = 23)	Macro dice (n = 213)
U-Net	0.723	0.496	0.607	0.554	0.624
U-Net++	0.727	0.497	0.590	0.538	0.625
DeepLabV3	0.742	0.524	0.623	0.574	0.650
DeepLabV3+	0.734	0.526	0.608	0.613	0.641
nnU-Net	0.767	0.528	0.624	0.585	0.665
TransUNet	0.737	0.428	0.526	0.534	0.599

Bold values indicate the highest (best) segmentation performance in each column.