Purpose: Nowadays, PET and dynamic contrast enhanced CT or MRI are used to assess tumor blood perfusion. Although [(15)O]H2O PET is the gold standard, it is hardly available for routine clinical practice, due to the short half-life of (15)O. However, the lack of uniformity in scanning and analytic methods limits the use of CT perfusion (CTP) in clinical trials and practice. This study compares [(15)O]H2O PET with CT based perfusion in lung tumors and assesses the effects of various CTP postprocessing and analytical methods on the CTP results using [(15)O]H2O PET as the reference technique.
Methods: Various CTP analysis and image postprocessing methods were assessed. Furthermore, parametric images were obtained using the Slope method. Volumes of interests were defined using several different segmentation methods including Hounsfield unit based contouring thresholds, both with and without framewise application of dynamic contouring thresholds to exclude lung tissue or intravascular contrast. A head-to-head comparison of tumor perfusion obtained by CTP and [(15)O]H2O PET was performed using linear regressions, Bland-Altman plots, and an intraclass correlation coefficient (ICC). In addition, the different postprocessing methods were compared reciprocally.
Results: In six lung cancer patients, perfusion assessed using CTP studies combined with the Slope method correlated best with [(15)O]H2O PET (ICC = 0.88; R(2) = 0.89; Y = 0.80). The Mullani-Gould method showed best correlation with the Slope method (ICC ≥ 0.71; R(2) ≥ 0.80; Y = 0.71-1.35). These correlations were obtained using dynamic contouring thresholds and show the influence of CTP postprocessing methods.
Conclusions: Tumor perfusion assessed by CTP in combination with dynamic contouring thresholds using the Slope method correlates well with [(15)O]H2O PET. This suggests that CTP can be used as a method to evaluate tumor perfusion in lung cancer.