TABLE 4.

Number of Independent Features per Segmentation Method, Number of Included Features, and Predictive Value for Largest Lesion for All Extracted Features (n = 483) and All Reliable, Repeatable, and Reproducible Features (n = 99)

No. of features	Parameter	Independent of MTV	Independent of SUVpeak	Independent of MTV and SUVpeak	Independent of MTV and SUVpeak and uncorrelated	No. of features in linear regression	CV-AUC (±SD)
483	SUV4.0	427	134	85	24	11	0.73 ± 0.10
	41%max	409	158	84	19	10	0.71 ± 0.11
	A50P	424	176	117	21	8	0.71 ± 0.11
	MV2	435	141	93	21	3	0.71 ± 0.10
	MV3	424	173	114	21	5	0.69 ± 0.11
	Best	437	168	122	25	10	0.69 ± 0.11
99	SUV4.0	57	46	13	10	5	0.73 ± 0.10
	41%max	50	57	14	6	1	0.65 ± 0.11
	A50P	54	59	20	9	4	0.63 ± 0.11
	MV2	59	52	18	8	3	0.70 ± 0.11
	MV3	54	52	18	7	3	0.67 ± 0.11
	Best	59	55	21	10	3	0.69 ± 0.11

41%max = 41% of SUV_max; A50P = 50% of SUV_peak; MV2 = majority vote segmenting voxels detected by ≥2 methods; MV3 = majority vote segmenting voxels detected by ≥3 methods.