Author(s): Duarte PS, Zhuang H, Castellucci P, Alavi A
Abstract Share this page
Abstract PURPOSE: The aim of this work was to determine the standard uptake value (SUV) threshold for differentiating malignant from benign bone lesions. MATERIAL AND METHODS: Ninety-nine bone sites in 33 patients who had undergone a 2-deoxy-2-[18F]fluoro-D-glucose-positron emission tomography (FDG-PET) study for cancer evaluation were studied. In addition to FDG-PET, a bone scan and at least two of the following determinations: magnetic resonance imaging (MRI), computed tomography (CT), and x-ray were conducted in each patient. The bone lesions were considered positive for malignancy if confirmed by clinical follow-up or a high degree of suspicion based on the positive results of at least three (which must include bone scan) out of four other imaging modalities. By these criteria, 39 lesions were considered positive and 60 were considered negative. The SUV values were classified as positive or negative using 61 different values of threshold (range from 1.0 to 7.0). These results were compared with the positive criteria above and reclassified as true positive, true negative, false positive, and false negative. The true-positive fraction and false-positive fraction were calculated for each threshold value. The receiver operating characteristic (ROC) curve was drawn and the best value was determined by visual analysis. RESULTS: The SUV threshold was considered 2.5. Twenty-nine out of 39 bone lesions classified as positive showed a SUV > 2.5. Of the 10 false-negative lesions, seven showed a SUV between 1.1 and 2.0, and three were not detected. Fifty-six out of 60 lesions classified as negative showed a SUV < 2.5. Four lesions were false positive: one was a rib fracture and three were severe degenerative changes in the lumbar spine. Using an SUV threshold of 2.5, the sensitivity was 74.3\% and the specificity was 93.3\%. CONCLUSION: In our patient population, the optimal SUV to classify a bone lesion as malignant or benign is 2.5.
This article was published in Mol Imaging Biol
and referenced in