Assessment of pancreatic tumor resectability with multidetector computed tomography: semiautomated console-generated images versus dedicated workstation-generated images.

RATIONALE AND OBJECTIVES

The purpose of this retrospective study was to compare the maximum intensity projection (MIP) images generated at a multidetector computed tomography (MDCT) scanner console using advanced tools at a three-dimensional (3D) workstation for assessment of pancreatic tumor resectability.

MATERIALS AND METHODS

Institutional review board approval and informed consent wavier were obtained for this retrospective study. The intraoperative findings were used as reference standard. Two radiologists assessed console MIPs that were created using computed tomographic (CT) data sets of 30 patients (17 men and 13 women; age range, 35-79 years; mean age, 58 years) operated for pancreatic tumors. Semi-automated MIP images were created on a separate MDCT console. Two blinded radiologist (R1, R2) and surgeons (S1, S2) evaluated the image data independently for vascular involvement and tumor resectability. The image quality and diagnostic confidence for MIPs were graded on a 5-point scale (1 = poor, 2 = suboptimal, 3 = intermediate, 4 = good; 5 = excellent) and comparison was made with 3D workstation image scores.

RESULTS

The findings revealed greater than 90% sensitivity, specificity, and accuracy for detecting involvement of peripancreatic vessels by pancreatic tumor with an excellent interobserver agreement (kappa = 0.87-1.00). The findings of console-generated MIPs were same as the findings of 3D workstation images. The mean of image quality and diagnostic confidence grading for console MIPs by assessors were 4.4 and 4.2, respectively. The average time to generate simple MIPs at the console was 3.4 minutes (range, 2.3-4.4) compared to 26 minutes (range, 18-33) to create images at the 3D workstation.

CONCLUSION

Semi-automated MIPs generated from an MDCT scanner console is an excellent alternative to 3D workstation images for assessing resectability of pancreatic tumor based on vascular involvement. Console MIPs can be quickly generated during the time of scan and thus can improve CT workflow.