Pixel-Label-Based Segmentation of Cross-Sectional Brain MRI Using Simplified SegNet Architecture-Based CNN

Table 3

Comparison of performance parameters for each result image (Figure 5(g)), with respective ground truth image (Figure 5(b)).

Test image ID	Parameter	CSF part	Gray part	White part	Mean value
OAS1_0081_MR1	Dice similarity	0.54	0.75	0.85	0.71
	Jaccard similarity	0.37	0.59	0.74	0.57
	Mean squared error	—	—	—	29.47
OAS1_0083_MR1	Dice similarity	0.84	0.75	0.79	0.80
	Jaccard similarity	0.73	0.60	0.66	0.66
	Mean squared error	—	—	—	19.32
OAS1_0084_MR1	Dice similarity	0.85	0.71	0.78	0.78
	Jaccard similarity	0.74	0.55	0.64	0.64
	Mean squared error	—	—	—	25.02
OAS1_0085_MR1	Dice similarity	0.72	0.67	0.73	0.71
	Jaccard similarity	0.56	0.51	0.57	0.55
	Mean squared error	—	—	—	32.52
OAS1_0086_MR1	Dice similarity	0.74	0.85	0.92	0.84
	Jaccard similarity	0.59	0.74	0.85	0.73
	Mean squared error	—	—	—	9.52
OAS1_0087_MR1	Dice similarity	0.64	0.75	0.85	0.74
	Jaccard similarity	0.47	0.60	0.74	0.60
	Mean squared error	—	—	—	27.58