Inter-modality variability of three-dimensional pancreatic volume measurements: CT versus MRI

To investigate the variabilities in pancreatic volume measurements using CT and MRI for the normal pancreas. This study included 21 participants with normal pancreases. All underwent both contrast-enhanced CT and MRI within 30 days (mean interval: 2.2 ± 5.1 days). Pancreatic segmentations based on water-phase and out-of-phase Dixon MRI images (3 mm) and portal venous phase CT images (1 mm and 3 mm) were performed. CT images (1 mm) and out-of-phase MRI images (3 mm) were segmented twice. Bland–Altman analyses, intraclass correlation coefficients (ICCs), and paired samples t-tests were used to compare pancreatic volumes from CT and MRI. The pancreatic volumes measured by MRI were 18.8% smaller than those from CT scans (MRI: 68.9 ± 22.6 cm3; CT: 84.8 ± 30.1 cm3; P < 0.001). For repeat measurements, CT (1 mm) and MRI (3 mm out-of-phase) showed excellent agreement, with CT yielding a mean volume difference of 9.7 cm3 and limits of agreement (LOAs) of − 7.5 to 27.0 cm3 (ICC: 0.919; Dice: 0.83), and MRI showing a mean volume difference of − 5.5 cm3 and LOAs of − 19.3 to 8.4 cm3 (ICC: 0.920; Dice: 0.80). No significant differences were found in pancreatic volumes measured by CT with 1-mm and 3-mm slice thickness (P = 0.07) or between water-phase and out-of-phase MRI (P = 0.60). The pancreatic volume measured by MRI was smaller than that measured by CT; it is not recommended to mix the CT and MRI if the follow-up of pancreatic volume is necessary during the therapy in the same patient.