Individually adapted, interactive multiplanar reformations vs. semi-automated coronary segmentation and curved planar...

To evaluate, whether semi-automated vessel extraction and curved planar reformations ("automated vessel extraction") increases diagnostic accuracy in the detection of relevant coronary artery lesions compared to manual, interactive multiplanar interpretation ("manual approach").

50 coronary CT angiography datasets were evaluated by four independent readers (two experienced, two novice) for the presence of stenoses exceeding 50% diameter reduction. One experienced and one novice reader each used the "manual approach" for cases 1-25 and "automated vessel extraction" for cases 26-50, while the other two readers used the complementary method. Results were compared to those of invasive coronary angiography.

Using the "manual approach", 37 of 42 stenoses were correctly detected by experienced as well as novice readers. 14 vs. 17 lesions were false positive (sensitivity 88%, specificity 91% vs. 89%, PPV 73% vs. 69%, NPV 97%, n.s.). Using "automated vessel extraction", experienced readers detected 35/42 stenoses compared to 31/42 for novice readers. 7 vs. 11 lesions were missed and 17 vs. 15 false-positive lesions reported (sensitivity 83% vs. 74%, specificity 89% vs. 90%, PPV 67%, NPV 95% vs. 93%, n.s.).In patient-based analysis, for novice readers sensitivity was higher using the "manual approach" (97%, 29/30 pts. vs. 80%, 24/30 pts., p=0.069).

Semi-automated vessel extraction and curved multiplanar reconstructions do not improve the diagnostic accuracy of coronary CT angiography compared to the use of interactive multiplanar reformations. Especially for less experienced readers, the use of automatically rendered curved multiplanar reconstructions alone cannot be recommended.