CT-perfusion measurements in pancreatic carcinoma with different kinetic models: Is there a chance for tumour grading based on functional parameters?

To evaluate the interchangeability of perfusion parameters obtained with help of models used for post-processing of perfusion-CT images in pancreatic adenocarcinoma and to determine the mean values and ranges of perfusion in different tumour gradings. Perfusion-CT imaging was performed prospectively in 48 consecutive patients with pancreatic adenocarcinoma. In 42 patients biopsy-proven tumor grading was available (4 × G1/24 × G2/14 × G3/6× unknown). Images were post-processed using a model based on the maximum-slope (MS) approach (blood flow-BFMS) + Patlak analysis (P) (blood volume [BVP] and permeability [k-transP]), as well as a model with deconvolution-based (D) analysis (BFD, BVD and k-transD). 50 mL contrast agent were applied with a delay time of 7 s. Perfusion parameters were compared using intraclass correlation coefficient (ICC), the Wilcoxon matched-pairs test and Bland-Altman plots. Forty eight VOIs of tumours were outlined and analysed. Moderate to good ICC values were found for the perfusion parameters (ICC = 0.62–0.75). Wilcoxon matched-pairs revealed significantly lower values (P< .001 and 0.008), for the BF and BV values obtained using the maximum-slope approach + Patlak analysis compared to deconvolution based analysis. For k-trans measurement, deconvolution revealed significantly lower values (P< 0.001). Different histologic subgroups (G1-G3) did not show significantly different functional parameters. There were significant differences in the perfusion parameters obtained using the different calculation methods, and therefore these parameters are not directly interchangeable. However, the magnitude of pairs of parametric values is in constant relation to each other enabling the use of any of these methods. VPCT parameters did not allow for histologic classification.