Optimized CT Attenuation and SUV Prediction Thresholds for Differentiating Enostoses From Untreated and Treated...

The objective of this study was to evaluate the accuracy of using CT attenuation and SUVs to differentiate enostoses from untreated and treated osteoblastic metastases on the attenuation-correction CT component of F-FDG PET/CTs.

We retrospectively reviewed F-FDG PET/CT studies of 117 patients (169 lesions), of which 65 had imaging of enostoses, and 52 had imaging showing the transition of lesions from untreated to treated osteoblastic metastases. We measured the mean CT attenuations and the SUVmax and SUVmean of each lesion. Receiver operating characteristic curve analyses were used to evaluate the accuracy of each metric in distinguishing enostoses from untreated and treated osteoblastic metastases.

For differentiating enostoses from untreated osteoblastic metastases, mean CT attenuation achieved an area under the receiver operating characteristic curve (AUC) of 90.8%, with an optimized threshold of 795 HU. SUVmax achieved an AUC of 94.9%, with an optimized threshold of 2.2. For differentiating enostoses from treated osteoblastic metastases, the AUCs for every metric decreased, with mean CT attenuation being the best at 82.7%. A joint predictive model combining both CT attenuation and SUV increased the AUC to 88.3%, and performance was significantly better than SUVmax or SUVmean alone (P = 0.029 and P = 0.049, respectively).

CT attenuation and SUV can reliably distinguish between enostoses and metastases on F-FDG PET/CT. However, the accuracy of these metrics decreases when used to differentiate enostoses from treated metastases. A joint prediction model combining CT attenuation with SUV can improve accuracy.