Table 3.

Statistical analysis: multiple comparisons for segmentation of PTC.

Measurement	(I) Method	(J) Method	Mean Diff. (I–J)	Std. Error	Sig.	95% C.I.
						Lo. Bound	Up. Bound
Accuracy	Proposed Method	U-Net	0.0784732 *	0.0068	$<0.001$	0.0651	0.0918
		SegNet	0.0761947 *	0.0068	$<0.001$	0.0629	0.0895
Precision	Proposed Method	U-Net	0.1187516 *	0.0289	$<0.001$	0.0620	0.1755
		SegNet	0.0452	0.0289	0.1180	−0.0116	0.1020
Recall	Proposed Method	U-Net	0.3336451 *	0.0271	$<0.001$	0.2803	0.3870
		SegNet	0.3856975 *	0.0271	$<0.001$	0.3323	0.4391
F1-score	Proposed Method	U-Net	0.2392282 *	0.0266	$<0.001$	0.1868	0.2916
		SegNet	0.2578479 *	0.0266	$<0.001$	0.2055	0.3102
Jaccard Index	Proposed Method	U-Net	0.3238887 *	0.0290	$<0.001$	0.2668	0.3809
		SegNet	0.3391651 *	0.0290	$<0.001$	0.2821	0.3962

* The proposed method is significantly better than the benchmark methods (U-net [27] and SegNet [28]) (p < 0.001).