System applies CAD to automate triage of chest pain patients

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon
 - Tech Heart

Triage of patients with acute chest pain may soon become easier, as application of a computer-aided simple triage (CAST) system for automatic stenosis detection can accurately identify patients with significant stenosis, according to a study published online June 2 in Academic Radiology.

“Given the significance of coronary artery disease as the most important socioeconomic health care problem in the Western world, the application of CAD [computer aided detection] algorithms and diagnosis techniques to this disease is surprisingly rare,” wrote Mathias Meyer, BSc, of University Medical Center Mannheim, Germany, and colleagues. “This study demonstrates that the cCTA [coronary CT angiography] CAST system evaluated in this study is a reliable tool to rule out significant stenotic coronary artery stenosis on a per-patient as well as on a per-vessel level and especially improves the diagnostic accuracy of an inexperienced reader in a consecutive cohort of patients with acute chest pain and an intermediate risk for ACS [acute coronary syndrome].”

Assessing patients who present to an emergency department with undefined chest pain is a challenge, though cCTA can reliably exclude significant coronary artery stenosis in patients with a low- to intermediate-risk profile for ACS, explained the authors. “However, a major limitation of cCTA for evaluation of chest pain patients in the ED is the lack of available experienced readers, especially during nighttime and weekend hours. Therefore, a CAD system with consistent performance for coronary artery stenosis detection appears desirable.”

Meyer and colleagues sought to assess a CAST system, which is a subclass of CAD used to perform initial triage, unlike conventional CAD systems which function primarily as a second reader. The study initially included 93 patients with acute chest pain and intermediate risk of ACS, among which 74 had adequate cCTA image quality for automatic analysis by the CAST system.

The CAST system detected stenosis of 50 percent or greater in 45 patients, compared with human expert interpretation, which identified 37 patients with such significant stenosis. On a per-patient level, the CAST system had a sensitivity and specificity of 100 and 78 percent, respectively. Per-vessel sensitivity and specificity was 79 and 89 percent, respectively.

The authors also examined the impact of CAST guidance on an inexperienced observer and found the system offered significant improvement. Sensitivity and positive predictive values for inexperienced readers in detecting significant stenosis rose from 69 and 41 percent, respectively, without CAST, to 91 and 74 percent, respectively, with CAST.

Meyer and colleagues noted that motion artifacts reduced image quality in some cases, caused vessels to be improperly identified and the presence of stenosis to be overcalled. Even in these cases, however, sensitivity and negative predictive value remained excellent, they noted.

“In addition, such CAST systems can be used to perform a reading order prioritization (eg, by giving higher priority to cases deemed positive by the system) or by assigning more experienced readers to positive or low quality cases and less experienced to simple negative cases.”