Deep learning-based automated segmentation of eight brain anatomical regions using head CT images in PET/CT

BMC Medical Imaging

Table 1 Dice scores of eight brain anatomical regions during the training and testing process with different deep learning models

Dice scores
	Basal ganglia		Cerebellum		Hemisphere		Hippocampus
	Left	Right	Left	Right	Left	Right	Left	Right
Training
DenseVNet 4000 epochs, LR = 0.001	0.944	0.932	0.939	0.919	0.963	0.961	0.872	0.864
DenseVNet 2000 epochs, LR = 0.001 2000 epochs, LR = 0.00025	0.943	0.937	0.938	0.917	0.963	0.960	0.888	0.854
3D U-Net	0.855	0.755	0.802	0.797	0.921	0.915	0.723	0.571
Testing
DenseVNet 4000 epochs, LR = 0.001	0.927	0.908	0.911	0.892	0.957	0.956	0.676	0.743
DenseVNet 2000 epochs, LR = 0.001 2000 epochs, LR = 0.00025	0.926	0.912	0.909	0.884	0.956	0.953	0.711	0.758
3D U-Net	0.891	0.745	0.808	0.812	0.866	0.839	0.544	0.565

ISSN: 1471-2342