Dynamic MDCT of the pancreas: Is time–density curve morphology useful for the differential diagnosis of solid lesions? A preliminary report

Abstract

CT perfusion has been proposed for pancreatic lesion characterization. However, scan and analysis protocols influence numerical data. To overcome this, the purpose of our study is to evaluate the use of time-density curves obtained from MDCT perfusion of the pancreas for the characterization of normal parenchyma, adenocarcinoma, chronic pancreatitis and endocrine tumors. 31 patients with solid pancreatic lesions and 21 patients with renal cell carcinoma underwent 64-row MDCT perfusion of the pancreas after injection of 50 cc of a 370 mg I/ml solution at 5 cc/s. 63 time-density curves were obtained from normal parenchyma (21 patients), adenocarcinoma (25), endocrine tumors (4) and atrophic parenchyma (13). Two readers independently categorized the 63 time-density curves into 4 different morphologies: normal wash-in and wash-out (A), low wash-in followed by plateau (B), low wash-in followed by faint wash-out (C) and high wash-in and wash-out (D). Interobserver agreement was calculated with kappa statistics. Fisher test was used to calculate sensitivity, specificity, positive (PPV) and negative (NPV) predictive values for each type of curve. Interobserver agreement was very good (Kappa=0.849). Curve A had 94.4% sensitivity, 91.1% specificity, 80.95% PPV, 97.6% NPV for 'normal parenchyma'. Curve B had 74.19% sensitivity, 93.75% specificity, 92% PPV, 78.95% NPV in diagnosing 'adenocarcinoma'. Curve C had 45.45% sensitivity, 84.62% specificity, 38.46% PPV, 88% NPV for 'chronic pancreatitis'. Curve D had 100% sensitivity, 98.33% specificity, 75% PPV, 100% NPV for 'endocrine tumor'. The morphology of MDCT perfusion time-density curves appears to be useful in characterizing pancreatic lesions, and might help overcome the differences in scan and postprocessing techniques.