Secretin-Enhanced Magnetic Resonance Cholangiopancreatography for Assessing Pancreatic Secretory Function in Children.

OBJECTIVE

To assess the accuracy and interrater reproducibility of measurements of pancreatic secretory function by magnetic resonance cholangiopancreatography in response to secretin administration and to describe our experience using the technique to noninvasively assess pancreatic secretory function in a pediatric population.

STUDY DESIGN

In the accuracy study, phantoms with varying fluid volume (47-206 mL) were imaged using the clinical quantification sequence. Fluid volume was measured by image segmentation (ImageJ). Measurement accuracy was expressed in terms of error (absolute and percent) relative to known fluid volume. In the reproducibility study and clinical experience, 31 patients with suspected pancreatic disease underwent 33 secretin-enhanced magnetic resonance cholangiopancreatography exams. Two-dimensional T2-weighted, fat-saturated single shot fast spin echo sequences were acquired before and after secretin injection (0.2 µg/kg, max 16 µg). Secreted fluid volume (postsecretin minus presecretin) was independently measured by 2 blinded reviewers. Between reviewer measurement reproducibility was assessed based on correlation (Spearman) and bias (Bland-Altman analysis).

RESULTS

For the accuracy study, fluid volumes were measured with mean volume errors of -0.3 to +12.5 mL (percent error -0.03% to +9.0%). For the reproducibility study, the mean secreted fluid volumes measured by reviewer 1 and reviewer 2 were 79.1 ± 54.3 mL (range 5.5-215.4) and 77.2 ± 47.1 mL (range 6.7-198.1 mL), respectively. Measured secreted fluid volumes were very strongly correlated (r = 0.922) between reviewers with a bias of only 1.9 mL (95% limits of agreement -40.5 to 44.2).

CONCLUSIONS

Measurement of fluid volume by magnetic resonance imaging is highly accurate with <10% (<13 mL) error in measured volume. Measurements of pancreatic secreted fluid volume in response to secretin by magnetic resonance cholangiopancreatography are highly reproducible with a bias of <2 mL between reviewers.