Skip to main content

View full-text article in PMC

. 2021 Aug 10;11:16244. doi: 10.1038/s41598-021-95545-y

Table 4.

Multiple comparisons for the segmentation of HSILs or higher (SQCC): Fisher’s LSD test.

LSD multiple comparisons
Dependent variable	(I) Method	(J) Method	Mean difference (I-J)	Std. error	Sig.	95% CI
Dependent variable	(I) Method	(J) Method	Mean difference (I-J)	Std. error	Sig.	Lower bound	Upper bound
Precision	Proposed method	U-net¹⁹	0.77696^a	0.09093	< 0.0001	0.5966	0.9573
Precision	Proposed method	SegNet²⁰	0.67273^a	0.09146	< 0.0001	0.4913	0.8541
Recall	Proposed method	U-net¹⁹	0.19469^a	0.04242	< 0.0001	0.1091	0.2803
Recall	Proposed method	SegNet²⁰	0.02283	0.04242	0.593	− 0.0628	0.1084
F-measure	Proposed method	U-net¹⁹	0.71301^a	0.09314	< 0.0001	0.5283	0.8977
F-measure	Proposed method	SegNet²⁰	0.61492^a	0.09369	< 0.0001	0.4291	0.8007
Jaccard	Proposed method	U-net¹⁹	0.71348^a	0.07790	< 0.0001	0.5590	0.8680
Jaccard	Proposed method	SegNet²⁰	0.61008^a	0.07836	< 0.0001	0.4547	0.7655

^aThe proposed method is significantly better than the benchmark methods (U-net¹⁹ and SegNet²⁰) ( $P < 0.0001$ ).