Abstract
2317
Objectives The present study aimed to determine whether 18FDG PET/CT with fractal analysis can help to differentiate malignant from benign pulmonary nodules.
Methods Data from 54 patients with suspected non-small cell lung cancer (NSCLC) determined by 18FDG PET/CT were retrospectively analyzed. Thirty-five nodules were confirmed as NSCLC and 19 were inflammatory lesions. The maximum standardized uptake value (SUV) and density fractal dimension (FD) in each target nodule were calculated from PET images. The FD is a quantitative index of tracer uptake heterogeneity, with a higher FD corresponding to increased heterogeneity. The FD was calculated by relating the logarithms of each cutoff of pixel radioactivity to the total number of pixels. The diagnostic accuracy of SUV and FD and the effect of nodule size on diagnostic accuracy were compared. We also compared the accuracy of CT alone with that of simultaneous CT and PET that shows the additional effects of 18FDG uptake distribution.
Results We decided based on ROC analysis using different SUV and FD threshold cutoffs, that an SUV of 4.24 and an FD of 0.0267 were optimal for differentiating malignant from benign pulmonary nodules using our equipment. The diagnostic accuracy of SUV and FD on PET was 68.5% (37/54) and 76.9% (40/52), respectively. Nodule size significantly correlated with SUV, but not with FD. Accuracy improved from 64.8% (35/54) for CT alone to 94.4% (51/54) for CT+PET with FD.
Conclusions The FD was significantly higher in benign, than in malignant nodules. Our findings suggest that density fractal analysis of 18FDG PET/CT is useful for the differential diagnosis of malignant and benign pulmonary nodules