Magnetic resonance cholangiopancreatography at 3 Tesla: Image quality comparison between 3D compressed sensing and 2D single-shot acquisitions

Objectives: To compare the image quality between compressed sensing (CS) 3D-magnetic resonance cholangiopancreatography (MRCP) using respiratory-triggered (RT) and breath-hold (BH) acquisitions and 2D single-shot breath-hold (SSBH) MRCP at 3 T MRI. Methods: 53 datasets were retrospectively assessed. 3D-MRCP with CS (RT-CS10, BH-CS24) and 2D-SSBH MRCP were acquired. Overall image quality, blurring/motion artifacts and discernibility of the pancreaticobiliary tree (PBT) structures were scored on a 4-point scale by 2 radiologists. The contrast ratio between the common bile duct and its adjacent tissue was measured by region-of-interest (ROI) analysis. Signal intensity increase at the boundaries of the ducts was quantified by line profiles to objectively characterize blurring and motion artifacts. Results: Total scan duration was 17 s for BH-CS24, 1m12 s for 2D-SSBH and 3m48 s for RT-CS10. Images acquired with CS were consistently rated superior in terms of image quality, background suppression, blurring and discernibility of PBT structures compared to 2D-SSBH images. RT-CS10 was superior to BH-CS24 for all ratings except for blurring. Objective analysis yielded the highest contrast ratio for RT-CS10 (0.91 +/- 0.04) followed by BH-C24 (0.88 +/- 0.05) and 2D-SSBH (0.85 +/- 0.06); one-way ANOVA P < 0.0001. The line-profile slope through the CBD was significantly higher in BH-CS24 (37.91 +/- 6.38% of maximum intensity/mm) compared to RT-CS10 (29.46 +/- 8.17% of maximum intensity/mm) and on par with 2D-SSBH (35.8 +/- 12.30% of maximum intensity/mm); one-way-ANOVA P = 0.017. Conclusion: CS allows acquisition of volumetric image data with improved image quality compared to SSBH. CS24 yields substantial gains in acquisition speed while robust towards artifacts, enabling diagnostic image quality with a single breath-hold acquisition.