Multislice CT coronary angiography: evaluation of an automatic vessel detection tool

To investigate the potential of a new detection tool for multislice CT (MSCT) coronary angiography with automatic display of curved multiplanar reformations and orthogonal cross-sections.

Thirty-five patients were consecutively enrolled in a prospective intention-to-diagnose study and examined using a MSCT scanner with 16 x 0.5 mm detector collimation and 400 ms gantry rotation time (Aquilion, Toshiba). A multisegment algorithm using up to four segments was applied for ECG-gated reconstruction. Automatic and manual detection of coronary arteries was conducted using the coronary artery CT protocol of a workstation (Vitrea 2, Version 3.3, Vital Images) to detect significant stenoses (> or = 50 %) in all segments of > or = 1.5 mm in diameter. Each detection tool was used by one reader who was blinded to the results of the other detection method and the results of conventional coronary angiography.

The overall sensitivity, specificity, nondiagnostic rate, and accuracy of the automatic and manual approach were 90 vs. 94 %, 89 vs. 84 %, 6 vs. 6 %, and 89 vs. 88 %, respectively (p = n. s.). The vessel length detected with the automatic and manual approach were highly correlated for the left main/left anterior descending (143 +/- 30 vs. 146 +/- 24 mm, r = 0.923, p < 0.001), left circumflex (94 +/- 35 vs. 93 +/- 33 mm, r = 0.945, p < 0.001), and right coronary artery (145 +/- 36 vs. 144 +/- 37 mm, r = 0.925, p < 0.001). The time required to create reformations along the coronary arteries was significantly shorter with the automatic tool compared to the manual approach (203 +/- 77 vs. 391 +/- 104 sec, p < 0.005). In 90 % of the coronary branches automatic detection required less time than the manual approach.

Automatic coronary vessel detection is feasible and reduces the time required to create reformations by a factor of approximately two without deteriorating the diagnostic accuracy.