Table 2. The results of Fisher linear classifier and Permutation for different combinations of features.

	Classification error	Sensitivity	Specificity	P (eact)	P (senact)	P (specact)
SUVmax and THKmax	0.21 (0.05)	0.82 (0.07)	0.63 (0.16)	0.014 (0.03)	0.050 (0.07)	0.295 (0.20)
SUVmax only	0.19 (0.04)	0.85 (0.06)	0.63 (0.17)	0.002 (0.01)	0.002 (0.01)	0.333 (0.18)
SUVmax/THKmax and THKmax	0.20 (0.04)	0.83 (0.06)	0.66 (0.17)	0.007 (0.02)	0.040 (0.05)	0.260 (0.19)
SUVmax/THKmax only	0.17 (0.03)	0.86 (0.05)	0.65 (0.14)	0.0002 (0.0009)	0.0002 (0.0009)	0.321 (0.18)

Data are mean (SD); P (e_act), P (sen_act) and P (spec_act) representing the mean p-value of the permutation test for the actual classification error, sensitivity and specificity.