Variability of ¹⁸F-FDG-positive lung lesion volume by thresholding

To assess the variability of (18)F-FDG-positive volume measurements in lung cancer patients, obtained with different fixed percentages of maximum standard uptake value (SUVmax) thresholds.

PET dynamic acquisition involving ten frames was performed within 60-110 min post-injection in eight patients. In each lesion (n = 11), volume was automatically outlined in each frame with fixed 40-50-60-70 % of the SUVmax thresholds. Thus, ten volume values for each threshold (V40-50-60-70) were available to calculate relative SD (SDr), and hence relative measurement error (MEr) and repeatability (R). Dependence on SUVmax variability was also assessed.

Mean SDr (; %) of volume estimates was found to strongly correlate with threshold value (T; %):  = 1.626 × exp(0.044 × T) (r = 0.999; P < 0.01). MEr and R for V40 were found to be (95 % CL) 18.9 % and 26.7 %. For all fixed thresholds, in successive frames of an arbitrary lesion, volume estimate inversely correlated with SUVmax (P ≤ 0.02).

A formula allows estimation of the variability of (18)F-FDG-positive volumes provided by any fixed percentage of SUVmax threshold, and hence by any thresholding method. It only necessitates conversion of the threshold value into the SUVmax percentage in order to aid quick estimation of volume variability magnitude in current clinical practice.