Multigrid Nonlocal Gaussian Mixture Model for Segmentation of Brain Tissues in Magnetic Resonance Images

Table 1

The average JS values (mean ± standard deviation) of GM, WM, and CSF segmentation obtained by applying four algorithms to T1-weighted brain MR images with increasing level of noise.

Algorithm	Tissues	3%	5%	7%	9%
GMM	WM	0.8512 ± 0.051	0.7532 ± 0.047	0.6574 ± 0.064	0.6135 ± 0.067
	GM	0.8478 ± 0.059	0.7231 ± 0.065	0.6326 ± 0.046	0.6012 ± 0.056
	CSF	0.8547 ± 0.048	0.7447 ± 0.054	0.6236 ± 0.043	0.6103 ± 0.055
Wells	WM	0.9201 ± 0.071	0.7932 ± 0.069	0.7154 ± 0.061	0.6843 ± 0.062
	GM	0.9102 ± 0.056	0.7863 ± 0.048	0.7001 ± 0.044	0.6632 ± 0.051
	CSF	0.8842 ± 0.052	0.7731 ± 0.059	0.7011 ± 0.061	0.6691 ± 0.058
MCFC	WM	0.9382 ± 0.051	0.8131 ± 0.053	0.7320 ± 0.046	0.6914 ± 0.038
	GM	0.9262 ± 0.048	0.7914 ± 0.046	0.7250 ± 0.047	0.6853 ± 0.032
	CSF	0.8937 ± 0.046	0.7724 ± 0.055	0.7123 ± 0.059	0.6749 ± 0.054
MNGMM	WM	0.9328 ± 0.019	0.9257 ± 0.032	0.9231 ± 0.031	0.9105 ± 0.032
	GM	0.9331 ± 0.017	0.9216 ± 0.038	0.9187 ± 0.034	0.9073 ± 0.038
	CSF	0.9293 ± 0.022	0.9211 ± 0.039	0.9127 ± 0.028	0.9005 ± 0.041