CT vs MRCP: Optimal classification of IPMN type and extent

Joshua A. Waters, C. Max Schmidt, Jason W. Pinchot, Patrick B. White, Oscar W. Cummings, Henry A. Pitt, Kumar Sandrasegaran, Fatih Akisik, Thomas J. Howard, Attila Nakeeb, Nicholas J. Zyromski, Keith D. Lillemoe

Research output: Contribution to journalArticle

151 Scopus citations


Intraductal papillary mucinous neoplasms (IPMNs) of the pancreas are being diagnosed with increased frequency. CT scanning commonly serves as the primary imaging modality before surgery. We hypothesized MRCP provides better characterization of IPMN type/extent, which more closely matches actual pathology. Of 214 patients treated with IPMN (1991-2006), 30 had both preoperative CT and MRCP. Of these, 18 met imaging study criteria. Independent readers performed retrospective, blinded analyses using standardized criteria for IPMN type and extent. A ductal connection was detected on 73% of MRCP scans and only 18% of CT. IPMN type was classified differently in seven (39%); four (22%) of which were read on CT as having main duct involvement where this was not appreciated on MRCP or found on surgical pathology. MRCP showed multifocal disease in 13(72%) versus only 9(50%) on CT. A different disease distribution was seen in 9(50%). Finally, 101 branch lesions were identified on MRCP compared to 46 on CT. CT falls short of MRCP in detecting a ductal connection, estimating main duct involvement, and identification of small branch duct cysts. These factors influence diagnostic accuracy, cancer risk stratification and operative strategy. MRCP should be employed for optimal management of patients with IPMN.

Original languageEnglish (US)
Pages (from-to)101-109
Number of pages9
JournalJournal of Gastrointestinal Surgery
Issue number1
StatePublished - Jan 1 2008


  • Computed tomography
  • Intraductal papillary mucinous neoplasm
  • Magnetic resonance cholangiopancreatography

ASJC Scopus subject areas

  • Surgery

Fingerprint Dive into the research topics of 'CT vs MRCP: Optimal classification of IPMN type and extent'. Together they form a unique fingerprint.

  • Cite this