Localized fine-tuning and clinical evaluation of deep-learning based auto-segmentation (DLAS) model for clinical target volume (CTV) and organs-at-risk (OAR) in rectal cancer radiotherapy

Table 2 Summary and statistical analysis of OAR structures (bladder, femoral heads, and small intestine) predicated by VPM and LFT in comparison with GT in metrics of DSC, 95HD, sensitivity and specificity

	DSC		95HD (mm)		Sensitivity		Specificity
	ExVal	GenEva	ExVal	GenEva	ExVal	GenEva	ExVal	GenEva
Bladder
VPM	0.967 ± 0.013	0.964 ± 0.015	2.837 ± 1.465	3.311 ± 1.723	0.972 ± 0.021	0.987 ± 0.017	0.996 ± 0.010	0.997 ± 0.08
LFT	0.961 ± 0.011	0.961 ± 0.014	3.327 ± 1.065	3.582 ± 1.881	0.965 ± 0.020	0.957 ± 0.028	0.992 ± 0.013	0.992 ± 0.017
p-value	< 0.01	< 0.01	0.32	0.51	0.02	< 0.01	0.03	< 0.01
Femoral heads
VPM	0.956 ± 0.013	0.972 ± 0.011	2.193 ± 0.573	1.922 ± 0.736	0.941 ± 0.019	0.972 ± 0.013	0.999 ± 0.001	0.999 ± 0.001
LFT	0.949 ± 0.008	0.970 ± 0.013	2.478 ± 0.532	1.978 ± 0.515	0.961 ± 0.012	0.976 ± 0.012	0.998 ± 0.001	0.998 ± 0.001
p-value	0.01	0.14	0.03	0.62	< 0.01	0.03	< 0.01	< 0.01
Small intestine
VPM	0.838 ± 0.072	0.806 ± 0.068	8.312 ± 2.821	11.605 ± 3.153	0.779 ± 0.121	0.723 ± 0.097	0.992 ± 0.005	0.993 ± 0.005
LFT	0.853 ± 0.051	0.834 ± 0.084	8.201 ± 3.229	10.888 ± 3.430	0.835 ± 0.096	0.808 ± 0.121	0.980 ± 0.017	0.985 ± 0.012
p-value	0.13	< 0.01	0.91	0.03	< 0.01	< 0.01	< 0.01	< 0.01

ISSN: 1748-717X