A research team from Heidelberg University found that CAD could yield a 100% negative predictive value in finding stenosis in emergency department (ED) patients with an intermediate pretest risk of acute coronary syndrome (ACS). CAD was also found to significantly improve the performance of an inexperienced radiologist.

"[CAD] holds great potential for improved workflow in cardiac imaging," said Dr. Thomas Henzler. He presented the study results during a scientific session on Saturday.

Believing that CAD may help to address the explosion in imaging data and studies, the researchers sought to evaluate the diagnostic accuracy of a commercially available CAD system for detecting coronary artery stenosis in CCTA. The retrospective study included 93 consecutive ED patients with acute chest pain and an intermediate pretest likelihood for ACS based on Thrombolysis in Myocardial Infarction (TIMI) criteria. All patients received CCTA.

The authors evaluated the performance of two expert radiologists (with five and 11 years of experience) as compared to the COR Analyzer (Rcadia Medical Imaging) CAD software. They also assessed the effect of CAD for an inexperienced reader with only four months of experience with CCTA.

The researchers defined stenosis as 50% or greater vessel diameter reduction in the left main (LM), left anterior descending (LAD), left circumflex (LCx), right coronary artery (RCA), or arterial side branches.

There were 48 men and 45 women in the study; the average patient age was 59. In 19 of 93 patients, the CAD software failed to provide results due to technical limitations. This was mainly due to motion artifacts (six cases), poor contrast enhancement (five patients), and patients with pacemakers (eight cases), Henzler said.

"However, all of those failed studies were reported as [having] limited diagnostic accuracy by our initial clinical report because we were not absolutely sure [that we could] rule out significant stenosis in those patients," he said.

CAD performance in comparison with expert readers was as follows:

• Sensitivity, per patient: 100%
• Sensitivity, per vessel: 79%
• Specificity, per patient: 78%
• Specificity, per vessel: 89%
• Negative predictive value, per patient: 100%
• Negative predictive value, per vessel: 94%
• Positive predictive value, per patient: 82%
• Positive predictive value, per vessel: 65%

The high negative predictive value from CAD was the most important result, Henzler said.

"We have a very high negative predictive value looking at all different vascular territories and 100% negative predictive value -- if the study is absolutely accurate -- to rule out significant stenosis," he said. "In our eyes, this is very important because you need some training to read the coronary CT angiography study and this helps inexperienced readers."

In individual segments, CAD found one of three LM lesions, 22 of 23 LAD lesions, seven of 13 LCx lesions, eight of 11 RCA lesions, and five of seven arterial side branches, according to the researchers. They noted that false-positive findings from CAD were related primarily to overestimating stenotic lesions with less than 50% diameter and calcified vessels. There were five false-positive results in the LM, five in the LAD, 10 in the RCA, and four in the arterial branches, according to the team.

After a relatively poor diagnostic performance without the use of CAD, the inexperienced reader benefited significantly from CAD, Henzler said.

"[CAD] improves the diagnostic performance of inexperienced readers," he concluded.

CAD boosts new readers in diagnosing stenosis on CCTA, November 28, 2010

Inexperienced readers benefit most from CAD in CCTA, November 15, 2010

Cardiac CAD safely rules out coronary artery disease, June 1, 2010

CAD takes on arteries with coronary stenosis detection, November 19, 2009

CAD software may help rads deal with CCTA data explosion, November 9, 2009