Abstract 6110: Impact of Different Post-Processing Techniques on Inter-Observer Agreement for Coronary CTA. Results from a Multicenter Trial
Many large scale studies have proven the high sensitivity and negative predictive value of computed tomographic angiography (CTA) for the detection of relevant coronary artery disease using invasive coronary angiography (ICA) as standard of reference. The aim of the present study was to evaluate the agreement of different observers and institutions for the analysis of quality and stenosis detection in a given CTA dataset using different post-processing techniques. As a secondary aim also the influence of the agreement between different observers on sensitivity and specificity was assessed. Six international centers with broad experience in coronary CTA submitted 41 CTA and corresponding ICA datasets for blinded analysis. The CTA datasets were randomly distributed to the participating centers and each CTA study was independently reviewed by 2–4 observers regarding quality and presence of relevant coronary stenoses. As post processing techniques
only axial images [AO] or
all desired techniques (axial images, VRT, curved MPR, MIP)[AT] were applied.
The image quality was assessed using a 4 step scale (0=not evaluable, 1=with doubts, 2=evaluable, 3=excellent), whereas the stenoses were graded as present or absent. A weighted agreement index for image quality and stenosis detection was calculated. altogether 471 coronary artery segments were analyzed. The mean quality level was significantly lower in AO than in AT (2.0±0.5 vs. 2.3±0.5; [p<0.001]). Conversely, the agreement index was higher in AO than in AT (0.62±0.19 vs. 0.57±0.15[p<0.001]). For the stenosis detection the agreement index was not significantly different (0.91±0.17[AO]) vs. 0.91±0.18 [AT]; p=n.s.). Sensitivity and specificity analysis were significantly depending on the agreement index, however no significant difference could be observed regarding the different post-processing techniques. The perception of the overall image quality increased with additional post processing techniques but not the agreement index. Better sensitivity and specificity of CTA was correlated to a higher agreement index but not to the use of additional post-processing techniques.