Table 2.

Evaluation of the classification results using the confusion tables. Tables from top to bottom represent the results for the MsFCM, MFCM, and FCM methods, respectively. On each table, the top row indicates the ground truth and the left column indicates the classified results. The value in each cell is the percentage against the ground truth. For example, the table at the top represents the results from the MsFCM method. For the column “CSF”, 96.77% pixels of the true CSF were correctly classified as Class “CSF”; 3.23% were incorrectly classified as Class “GM”; and none pixel was classified as Class “WM”. Thus, the false negative value is 3.23%. The images are T1-weighted MR images from McGill brain database with 9% noise and 20% field inhomogeneity.

MsFCM	Ground Truth
		CSF	GM	WM	FP
Classified Results	CSF	96.77±0.93	6.68±1.70	0.00±0.00	3.08±0.77
	GM	3.23±0.93	87.22±2.36	5.38±1.27	4.86±0.94
	WM	0.00±0.00	6.10±2.10	94.62±1.27	4.33±1.15
FN		3.23±0.93	12.78±2.36	5.38±1.27
MFCM	Ground Truth
		CSF	GM	WM	FP
Classified Results	CSF	99.30±0.61	22.39±11.16	0.03±0.06	9.99±4.30
	GM	0.69±0.61	68.48±7.77	6.34±1.53	5.07±1.28
	WM	0.01±0.02	9.13±4.86	93.63±1.56	6.44±3.01
FN		0.70±0.61	31.52±7.77	6.37±1.56
FCM	Ground Truth
		CSF	GM	WM	FP
Classified Results	CSF	97.62±1.21	16.75±7.01	0.03±0.04	7.52±2.60
	GM	2.37±1.21	72.47±4.23	13.19±2.52	10.76±2.29
	WM	0.01±0.02	10.78±4.70	86.78±2.55	7.64±2.85
FN		2.38±1.21	27.53±4.23	13.22±2.55