TABLE 4.

Various SUV Definitions of Avril et al. with Their Receiver-Operating-Characteristic (ROC) Areas for Breast Cancer Diagnosis

Method of obtaining SUV*†				Benign (n = 46)		Malignant (n = 51)			ROC area
ROI choice	Partial volume	Correction		Benign (n = 46)		Malignant (n = 51)			ROC area
ROI choice	Partial volume	Glucose	Body fat	S̅U̅V̅	CV	S̅U̅V̅	CV	FOM	Formula	Curve fit†
Average	Yes	No	No	1.5	0.50	4.2	0.60	1.319	0.906	0.91 ± 0.03
Average	Yes	Yes	No	1.1	0.58	3.2	0.62	1.258	0.896	0.92 ± 0.03
Maximum	Yes	No	No	2.0	0.43	5.1	0.61	1.254	0.895	0.87 ± 0.04
Average	No	Yes	No	1.0	0.40	2.7	0.70	1.232	0.891	0.87 ± 0.04
Maximum	Yes	Yes	No	1.5	0.51	3.9	0.62	1.190	0.883	0.87 ± 0.04
Average	No	No	No	1.4	0.36	3.6	0.71	1.187	0.882	0.86 ± 0.04
Average	No	No	Surface area	3.8	0.35	9.2	0.66	1.184	0.882	0.87 ± 0.04
Average	No	No	Lean mass	1.3	0.37	3.0	0.65	1.118	0.868	0.85 ± 0.04
Maximum	No	Yes	No	1.4	0.36	3.3	0.70	1.090	0.862	0.85 ± 0.04
Maximum	No	No	No	1.9	0.32	4.3	0.70	1.061	0.856	0.81 ± 0.04
Average				1.5‡	0.42	3.7‡	0.66
CV of average				0.22‡		0.20‡

↵* PET region of interest (ROI) may use average or hottest pixels, possibly with corrections for partial-volume effect and departure from population averages for glucose plasma concentration or body surface area. Lean or total body mass might also be used in whole-body specific activity calculations.
↵† From Avril et al. (22).
↵‡ All methods except 1 based on surface area are combined.
FOM = figure of merit defined by Equation 2.