Interobserver agreement in the interpretation of computed tomography in acute pulmonary embolism

Abstract

Multidetector computed tomography (MDCT) is one of the best diagnostic tools for the diagnosis of pulmonary embolism (PE). However, differences in MDCT interpretation, depending on the operator personal expertise, is an important factor that could interfere with the right diagnosis and, consequently, with the more adequate and well-timed therapy. The aim of the present study was to evaluate the interobserver agreement in the interpretation of MDCT for the diagnosis of acute PE. On a blind basis, 4 radiologists with different expertise in CT interpretation evaluated 46 different MDCT executed for acute PE. They had to verify the presence or absence of PE and, in the positive case, localize (right-left) and quantify (massive, segmentarian or subsegmentarian) it. The interobserver concordance was expressed using the Cohen K statistic. The mean concordance between the 4 operators was high (0.82; range, 0.68-0.95). Ruling out the massive PE cases, the mean concordance over the other cases was only moderate (0.47; range, 0.16-0.84). We found a very good interobserver agreement in MDCT evaluation for the diagnosis of massive PE, whereas we observed a lower concordance in regard to segmentarian and subsegmentarian PE. In the case of negative or nonmassive PE diagnosis, a second evaluation of the CT performed by an expert CT radiologist would probably be effective to decrease the CT evaluation error.