RSNA 2003 Scientific Papers > Performance of Dual-phase, Thin-section Pancreatic ...
  Scientific Papers
  SESSION: Gastrointestinal (Pancreas: Cystic Neoplasm-CT, MR)

Performance of Dual-phase, Thin-section Pancreatic CT in Differentiating Surgically-proven Pancreatic Cystic Lesions

  DATE: Wednesday, December 03 2003
  START TIME: 03:30 PM
  END TIME: 03:37 PM
  LOCATION: Room E450B
  CODE: M08-1119

Ng Vuong
Los Angeles , CA
Steven Raman MD
Peter Zimmerman MD
James Sayre PhD
David Lu MD

Computed tomography (CT), thin-section
Pancreas, cysts
Pancreas, neoplasms

Purpose: To determine the ability of dual-phase thin-section pancreatic CT in differentiating pancreatic cystic lesions.

Methods and Materials: We studied 47 consecutive patients with dual phase, thin collimation (< 3mm) CT and 47 surgically & pathologically proven cystic pancreatic lesions. Two experienced abdominal imagers blinded to clinical information independently reviewed scans in random order and classified lesions initially as benign (true pancreatic cyst, pseudocyst or abscess) or neoplastic. Readers further subcategorized neoplastic lesions as serous tumor, mucinous tumor, indeterminate (serous vs. mucinous), or other cystic neoplasm. The readers then re-classified lesions when given history of pancreatitis in 6/47 patients. A 2-reader consensus reading was also performed. Lesion characteristics recorded include lesion size, location, cyst number, largest cyst size, wall thickness, presence & pattern of calcification (central or/and peripheral) and ductal dilatation based on consensus reading. A multivariate logistic regression analysis was performed to determine significance of these parameters in differentiating between serous and mucinous tumors.

Results: There were 38 neoplastic lesions (15 serous cystadenomas, 1 serous cystadenocarcinoma, 18 mucinous tumors (4 IPMT), 2 SPEN, 1 ovarian cystadenocarcinoma metastases, 1 lymphangiosarcoma) and 9 benign lesions (6 true cysts, 3 pseudocysts). On initial reading, readers 1, 2 and consensus correctly classified neoplastic vs. non-neoplastic lesions in 80%, 83%, and 83% respectively. With history of presence or absence of pancreatitis provided, accuracy increased to 87%, 87%, and 89% respectively. Reader accuracy for identifying serous tumors from all other lesions was 75% (R1), 77% (R2), 81% (C). This did not change with provision of pancreatitis history. For mucinous tumors, accuracy increased (R1, R2, C: 53%, 57%, 60% to 60%, 62%, 66%) after provision of history. Logistic regression showed number of cysts > 6 and cysts < 2 cm as the two most significant parameters in differentiating between serous and mucinous tumors. In addition, if other features are taken into account including age, sex, and peripheral calcifications, discrimination accuracy between serous and mucinous tumors reached 94%.

Conclusion: Pancreatic CT enables a confident diagnosis of pancreatic cystic neoplasms relative to benign cysts; however significant overlap in imaging characteristics between serous and mucinous tumors precludes high accuracy.