Purpose: To evaluate multislice-CT versus MRI in the diagnosis and staging of pancreatic carcinoma in a prospective multi-reader analysis.
Materials and methods: Fifty patients with suspected pancreatic carcinoma underwent both multislice-CT (4-Row, "hydro-technique") and state-of-the-art MRI (two 1.5 T units). In correlation with histopathologic findings or in case of a non-lesion diagnosis by follow-up (6-month interval), we evaluated MRI versus CT in a multi-reader analysis (2 reader pairs) for: 1. diagnostic quality; 2. examination time; 3. accuracy of potential resectability; 4. kappa analysis of observer variations; and 5. overall diagnostic reliability.
Results: A total of 28 lesions (n = 22 malignant, n = 6 benign) were present in the cohort group versus 22 patients without a focal lesion (n = 10 pancreatitis, n = 12 no tumor). For lesion detection, CT had a sensitivity of 100/89 % (reader pair 1/2) and specificity of 77 %, and MRI had a sensitivity of 75/89 % and specificity of 77/73 %. For the subgroup of adenocarcinomas of the pancreas (n = 17), we found a sensitivity of 100 % and a specificity of 61 % for CT versus a sensitivity of 82/94 % and a specificity of 67/61 % for MRI. The accuracy for determining the resectability was 91/82 % for CT and 90/82 % for MRI. The kappa analysis showed a good correlation for CT (0.71) and a moderate correlation of both groups for MRI (0.49).
Conclusion: CT and MRI showed comparable results in the detection of pancreatic carcinomas as well as in the determination of resectability. Chronic pancreatitis as a "tumor-like-lesion" was the major factor of a missed diagnosis. The results of multi-reader analysis for both reading groups were almost identical with a moderate to good kappa correlation. There is no reason to prefer MRI (more expensive) over CT for patients with the presumptive diagnosis of pancreatic cancer.