Diagnostic accuracy of intra-abdominal fluid collection characterization in the era of multidetector computed tomography.

The purpose of this study was to test the hypothesis that with improved technology, the presence of abscess in a postoperative fluid collection may be prospectively made. This is an Institutional Review Board-approved, Health Insurance Portability and Accountability Act-compliant retrospective review of computed tomography (CT)-guided intra-abdominal fluid collection drainages. The diagnostic CT scans of 95 consecutive patients performed on 16- and 64-multidetector CT (MDCT) were reviewed by four readers with varying abdominal imaging expertise. Readers were asked to determine fluid content, to document whether infection was present, and to rate reader confidence for infection. A fifth radiologist reviewed the collections for imaging characteristics. The gold standard for presence of infection was microbiological Gram stain and culture. The logistic regression model showed that both fluid collections containing gas or high attenuation fluid (average CT density 20 or greater Hounsfield units) are significant predictors of infection (P = 0.001). The average sensitivity over the four readers for determining presence of infection was 83.4 per cent and specificity was 39.3 per cent. Even in the era of MDCT, the ability to predict whether or not a fluid collection is infected or not, based on imaging findings alone, is limited. Presence of gas is a strong indicator of infection, but no imaging finding is characteristic of a sterile fluid collection.