Detection and characterization of traumatic bile leaks using Gd-EOB-DTPA enhanced magnetic resonance cholangiography

Yon Cheong Wong, Li Jen Wang, Cheng Hsien Wu, Huan Wu Chen, Chen Ju Fu, Kuo Ching Yuan, Being Chuan Lin, Yu Pao Hsu, Shih Ching Kang

Research output: Contribution to journalArticlepeer-review

12 Citations (Scopus)


Expanding bile leaks after blunt liver trauma require more aggressive treatment than contained bile leaks. In this retrospective study approved by institution review board, we analyzed if non-invasive contrast-enhanced magnetic resonance cholangiography (CEMRC) using hepatocyte-specific contrast agent (gadoxetic acid disodium) could detect and characterize traumatic bile leaks. Between March 2012 and December 2014, written informed consents from 22 included patients (17 men, 5 women) with a median age of 24.5 years (IQR 21.8, 36.0 years) were obtained. Biliary tree visualization and bile leak detection on CEMRC acquired at 10, 20, 30, 90 minutes time points were independently graded by three radiologists on a 5-point Likert scale. Intraclass Correlation (ICC) was computed as estimates of interrater reliability. Accuracy was measured by area under receiver operating characteristic curves (AUROC). Biliary tree visualization was the best on CEMRC at 90 minutes (score 4.30) with excellent inter-rater reliability (ICC = 0.930). Of 22 CEMRC, 15 had bile leak (8 expanding, 7 contained). The largest AUROC of bile leak detection by three radiologists were 0.824, 0.914, 0.929 respectively on CEMRC at 90 minutes with ICC of 0.816. In conclusion, bile leaks of blunt liver trauma can be accurately detected and characterized on CEMRC.

Original languageEnglish
Article number14612
JournalScientific Reports
Issue number1
Publication statusPublished - Dec 1 2018
Externally publishedYes

ASJC Scopus subject areas

  • General


Dive into the research topics of 'Detection and characterization of traumatic bile leaks using Gd-EOB-DTPA enhanced magnetic resonance cholangiography'. Together they form a unique fingerprint.

Cite this