Smart contours: deep learning-driven internal gross tumor volume delineation in non-small cell lung cancer using 4D CT maximum and average intensity projections

Table 2 Performance of deep-learning models for IGTV delineation using MIP images for the test cohort

	IGTV_MIP−manu	Attention U-net	U-net	V-net
DSC	0.713 ± 0.117	0.852 ± 0.053	0.829 ± 0.055	0.819 ± 0.073
Surface DSC	0.466 ± 0.231	0.760 ± 0.130	0.688 ± 0.150	0.673 ± 0.123
HD95 (mm)	4.146 ± 2.229	3.209 ± 2.136	3.264 ± 2.015	3.903 ± 2.157
Sensitivity	0.571 ± 0.151	0.818 ± 0.108	0.849 ± 0.095	0.759 ± 0.142
Specificity	1.000 ± 0.000	0.999 ± 0.001	0.999 ± 0.001	0.999 ± 0.001

Note: The IGTV_MIP−manu column presents the comparison between the manually contoured IGTV on MIP images and the gold standard IGTV (IGTV_gt), whereas the model columns present the comparison between the model predictions and the IGTV_gt. The model with the best performance in each metric is highlighted in bold

ISSN: 1748-717X