Standardized uptake value-based evaluations of solitary pulmonary nodules using F-18 fluorodeoxyglucose-PET/computed tomography.

OBJECTIVE

Combined positron emission tomography and computed tomography (PET/CT) might improve the accuracy of PET tracer quantification by providing the exact tumour contour from coregistered CT images. We compared various semiquantitative approaches for the characterization of solitary pulmonary nodules (SPNs) using F-18 fluorodeoxyglucose PET/CT.

METHODS

The final diagnosis of 49 SPNs (46 patients) was based on histopathology (n=33) or patient follow-up (n=16). The regions of interest (ROIs) were drawn around lesions based on the CT tumour contour and mirrored to the coregistered PET images. Quantification of F-18 fluorodeoxyglucose uptake was accomplished by calculating the standardized uptake value (SUV) using three different methods based on: activity from the maximum-valued pixel within the tumour (SUV-max); the mean ROI activity within the transaxial slice containing the maximum-valued pixel (SUV-mean); and the mean activity over the full tumour volume (SUV-vol). SUVs were corrected for partial volume effects and normalized by body surface area, lean body weight, and blood glucose. Recovery coefficients for partial-volume correction were derived from phantom studies. The ability of various SUVs to differentiate between benign and malignant SPNs was determined by calculating the area under the receiver operating characteristic (ROC) curves.

RESULTS

Twenty-six SPNs were malignant and 23 were benign. The area under the ROC curve was 0.78 for SUV-mean, 0.83 for SUV-max, and 0.78 for SUV-vol. SUV-max and its normalizations yielded the highest area under the ROC curve (0.83-0.85); SUV-mean-partial volume corrected-lean body weight resulted in the lowest area under the ROC curve (0.76). At a specificity of 80%, SUV-max-body surface area provided the highest sensitivity (81%) and accuracy (80%) to detect malignant SPN. Using SUV-max with a cutoff of 2.4 at a specificity of 80% resulted in a sensitivity of 62% (accuracy 71%).

CONCLUSION

Various normalizations applied to SUV-max provided the highest diagnostic accuracy for characterization of SPNs. Quantification methods using the exact tumour contour derived from CT in combined PET/CT imaging (ROI mean activity within a single transaxial slice and mean tumour volume activity) did not result in improved differentiation between benign and malignant SPN. Obtaining SUV-max might be sufficient in the clinical setting.