Page 158 - Quantitative Imaging of Small Tumours with Positron Emission Tomography

results of the SUVmax 2.5 threshold were used in meta-analysis (SUVmean for PVC data in Hickeson et al.) since this was reported in all 5 studies with predefined SUV thresholds. One study used data-driven thresholds specifically for uncorrected (SUV 2.4) and PVC data (SUV 2.9) (18). Pooled sensitivity and specificity of uncorrected data were 81% (95%CI 70-89) and 70% (95%CI 48- 86), respectively (Figure 7.5). Pooled sensitivity and specificity of partial-volume corrected data were 91% (95%CI 83-95) and 60% (95%CI 37-79), respectively (Figure 7.4). No significant change in diagnostic accuracy after PVC was found (p=0.222), using the SUV thresholds as described above. One of the pulmonary studies (by Hickeson et al.) stratified both uncorrected and corrected data for lesion size (15). The authors observed that for lesions <2cm accuracy increased from 59% to 85% using SUV cut-off 2.5, while for lesions >2cm accuracy changed from 95% to 100%. 7 Systematic review and meta-analysis Figure 7.3: Forest plots presenting sensitivity (a) and specificity (b) with 95% CI of discrimination between benign and malignant pulmonary nodules with 18F-FDG-PET. 157