Cancer prognosis using support vector regression in imaging modality

Table 1 Cross validation results for three feature selection methods using 32 features and 194 instances

	SVCR			Cox regression
	IS	FFS	NS	IS	FFS	NS
CI	0.6316	0.6333	0.6321	0.6623	0.6333	0.5677
BS	0.2770	0.3609	0.2675	0.2487	0.3319	0.2294

IS: Individual feature selection; FFS: Forward feature selection; NS: Non-feature selection; SVCR: Support vector regression method.

Table 2 Cross validation results for three feature selection methods and without lymph nodes status (32 features) and 198 instances

	SVCR			Cox regression
	IS	FFS	NS	IS	FFS	NS
CI	0.6532	0.6733	0.6579	0.6465	0.6733	0.5981
BS	0.1969	0.3468	0.2109	0.2504	0.4975	0.3321

IS: Individual feature selection; FFS: Forward feature selection; NS: Non-feature selection; SVCR: Support vector regression method.

Table 3 Cross validation results for three feature selection methods and stratified sampling with respect to tumor size and 194 instances

	SVCR			Cox regression
	IS	FFS	NS	IS	FFS	NS
CI	0.6390	0.6021	0.6246	0.6244	0.6021	0.5867
BS	0.2904	0.3149	0.3377	0.2588	0.4424	0.3390

IS: Individual feature selection; FFS: Forward feature selection; NS: Non-feature selection; SVCR: Support vector regression method.

Table 4 Cross validation results for three feature selection methods and stratified sampling with respect to lymph nodes status and 194 instances

	SVCR			Cox regression
	IS	FFS	NS	IS	FFS	NS
CI	0.6845	0.6769	0.6572	0.6341	0.6769	0.6145
BS	0.2065	0.3311	0.2643	0.2147	0.3400	0.3273

IS: Individual feature selection; FFS: Forward feature selection; NS: Non-feature selection; SVCR: Support vector regression method.