Brain Lesion Segmentation Based on Joint Constraints of Low-Rank Representation and Sparse Representation

Table 1

Comparison of segmentation accuracy obtained by different methods for two cases.

Lesions	Subjects	Number of training samples	Global-RX [35]	SRD [36]	LRR	JCLRRSR
Brain tumor	HGG	200	0.7542	0.6803	0.7016	0.7624
		500	0.7634	0.8153	0.8565	0.9175
		800	0.7689	0.8209	0.8602	0.9213
	LGG	200	0.7227	0.5422	0.6014	0.6624
		500	0.7272	0.7903	0.8325	0.8951
		800	0.7305	0.8023	0.8412	0.9031
	Total	200	0.7384	0.6112	0.6515	0.7148
		500	0.7503	0.8028	0.8445	0.9063
		800	0.7497	0.8116	0.8507	0.9122
Multiple sclerosis	BMSL	1000	0.6674	0.5213	0.5641	0.6425
		2000	0.6846	0.7235	0.7637	0.8026
		3000	0.6855	0.7321	0.7732	0.8242
	SMSL	1000	0.5326	0.4865	0.5245	0.6057
		2000	0.5578	0.6395	0.6835	0.7344
		3000	0.5587	0.6400	0.6910	0.7356
	Total	1000	0.6000	0.5039	0.5443	0.6241
		2000	0.6212	0.6815	0.7236	0.7685
		3000	0.6221	0.6860	0.7321	0.7799