Influence of ROI definition, partial volume correction and SUV normalization on SUV-survival correlation in oesophageal cancer

OBJECTIVE: An explanation for the discrepancies in the reported correlations between standardized uptake value (SUV) and survival might be the application of different SUV methodologies. The primary aim of this study was to examine the influence of using different methodologies on SUV-survival correlation. METHODS: Data were used from a prospective cohort study consisting of oesophageal cancer patients in whom preoperative fluorodeoxyglucose positron emission tomography was performed. Various methodologies of SUV calculation/correction were correlated with the default (SUV A41% corrected for body surface area): different volume of interest definitions, different SUV normalization, with and without serum glucose correction, and with (PVC+) and without partial volume correction (PVC-). Receiver operating characteristic (ROC) curves using any type of SUV for the identification of potential correlation with disease-free survival were also compared. RESULTS: Fifty-two patients were included for this study. Significant correlations were found between SUV A41% and all the other described SUVs: SUV 50% (r2=0.99; P<0.001), SUV A50% (2=0.98; P<0.001), SUVmax ( 2=0.98; P<0.001), SUV A41% PVC+ (2=0.97; P<0.001) and SUV A41% glucose (2=0.93; P<0.001). No correlation was found between volume of interest 41% and SUV A41%, with or without, PVC (P=0.85 and P=0.41). Significant correlations were found between SUVmax corrected for body surface area, SUVmax corrected for body weight (2=0.96; P<0.001) and SUV corrected for lean body mass (2=0.98; P<0.001). ROC curves for various SUV methodologies showed an almost identical area under the curve for any type of SUV. CONCLUSION: A strong correlation was found between all the investigated SUV methodologies. Moreover, when looking for correlations between SUV and disease-free survival, the areas under the ROC curves were almost identical for any type of SUV methodology.

• View at Publisher
• View at Pubmed
• View at Scopus
• View at Web of Science