Utility of a novel integrated deep convolutional neural network for the segmentation of hip joint from computed tomography images in the preoperative planning of total hip arthroplasty

Table 3 Multiple comparison of CMG Net and the order four nets of segmentation time

Disease	(I) NET	(J) NET	Mean difference (I − J)	Std. error	Sig	95% confidence interval
Disease	(I) NET	(J) NET	Mean difference (I − J)	Std. error	Sig	Lower bound	Upper bound
Multiple comparisons
Dependent variable: TIME
Least significance difference (LSD)
ONFH	CMG net	FCN	− 10.66	0.36	< 0.05	− 11.37	− 9.95
		2D UNET	− 9.87	0.36	< 0.05	− 10.58	− 9.16
		2.5D UNET	− 31.55	0.36	< 0.05	− 32.26	− 30.84
		3D UNET	− 41.33	0.36	< 0.05	− 42.04	− 40.62
DDH	CMG net	FCN	− 10.22	0.22	< 0.05	− 10.66	− 9.78
		2D UNET	− 10.15	0.22	< 0.05	− 10.59	− 9.71
		2.5D UNET	− 32.52	0.22	< 0.05	− 32.96	− 32.09
		3D UNET	− 40.69	0.22	< 0.05	− 41.13	− 40.25
FNF	CMG net	FCN	− 9.84	0.12	< 0.05	− 10.07	− 9.61
		2D UNET	− 9.81	0.12	< 0.05	− 10.04	− 9.58
		2.5D UNET	− 31.43	0.12	< 0.05	− 31.66	− 31.20
		3D UNET	− 39.93	0.12	< 0.05	− 40.16	− 39.69
OA	CMG net	FCN	− 8.63	0.32	< 0.05	− 9.25	− 8.00
		2D UNET	− 10.51	0.32	< 0.05	− 11.13	− 9.88
		2.5D UNET	− 29.66	0.32	< 0.05	− 30.29	− 29.04
		3D UNET	− 40.06	0.32	< 0.05	− 40.69	− 39.44
NORMAL	CMG net	FCN	− 8.97	0.32	< 0.05	− 9.59	− 8.34
		2D UNET	− 10.83	0.32	< 0.05	− 11.46	− 10.21
		2.5D UNET	− 30.33	0.32	< 0.05	− 30.96	− 29.70
		3D UNET	− 40.23	0.32	< 0.05	− 40.86	− 39.60

ISSN: 1749-799X