SCCT: Computer-aided analysis for CAD detection shows promise
Computer-aided analysis (CAA) showed high sensitivity and high negative predictive value for the evaluation of significant coronary artery disease (CAD) on 64-slice cardiovascular CT angiography (CCTA) across three populations with differing CAD prevalence, suggesting that CAA could be used in clinical practice, based on a study presented this week at the fifth annual meeting of the Society of Cardiovascular Computed Tomography (SCCT).
CCTA is increasingly used for the assessment of CAD in symptomatic patients. As a result, Melissa Daubert, MD, from the Stony Brook University Medical Center in Stony Brook, N.Y., and colleagues undertook the study to evaluate the clinical applicability of CAA software for the detection of significant coronary stenosis on 64-slice CCTA in three patient populations with low (8 percent), moderate (13 percent) and high (27 percent) CAD prevalence.
The researchers performed analysis on 341 consecutive patients with appropriate clinical indications for 64-slice CCTA at three clinical sites in the U.S. CAA software (COR Analyzer, Rcadia Medical Imaging in Haifa, Israel) performed automatic segmentation, tracking and detection of significant coronary lesions (defined as the most severe lesion in each segment with more than 50 percent stenosis) by finding the best match between extracted features and a large dataset of interpreted studies.
Two experienced cardiologists compared the CAA results to the consensus manual interpretation. The investigators conducted a data analysis on a per patient and per segment basis.
CAA per patient had a sensitivity of 100 percent across all three clinical sites, Daubert and colleagues reported. The specificity in the low, moderate and high CAD prevalence populations was 64 percent, 41 percent and 38 percent, respectively.
Likewise, the researchers reported that the negative predictive value at the three clinical sites was 100 percent. The positive predictive value was 22 percent, 21 percent and 38 percent for the low, moderate and high CAD prevalence populations, respectively. The segmental analysis yielded results with a similar trend.
The "results suggest that [CAA software] can be used in clinical practice to facilitate the accurate detection and exclusion of CAD on 64-slice and over CT scanners," said the study’s senior author Michael Poon, MD, director of advanced cardiac imaging in the department of radiology at Stony Brook. CAA software results "do not replace the expert analysis by the interpreting cardiologist, but rather work synergistically by combining the cardiologist's expertise with the software's capability," he said.
CCTA is increasingly used for the assessment of CAD in symptomatic patients. As a result, Melissa Daubert, MD, from the Stony Brook University Medical Center in Stony Brook, N.Y., and colleagues undertook the study to evaluate the clinical applicability of CAA software for the detection of significant coronary stenosis on 64-slice CCTA in three patient populations with low (8 percent), moderate (13 percent) and high (27 percent) CAD prevalence.
The researchers performed analysis on 341 consecutive patients with appropriate clinical indications for 64-slice CCTA at three clinical sites in the U.S. CAA software (COR Analyzer, Rcadia Medical Imaging in Haifa, Israel) performed automatic segmentation, tracking and detection of significant coronary lesions (defined as the most severe lesion in each segment with more than 50 percent stenosis) by finding the best match between extracted features and a large dataset of interpreted studies.
Two experienced cardiologists compared the CAA results to the consensus manual interpretation. The investigators conducted a data analysis on a per patient and per segment basis.
CAA per patient had a sensitivity of 100 percent across all three clinical sites, Daubert and colleagues reported. The specificity in the low, moderate and high CAD prevalence populations was 64 percent, 41 percent and 38 percent, respectively.
Likewise, the researchers reported that the negative predictive value at the three clinical sites was 100 percent. The positive predictive value was 22 percent, 21 percent and 38 percent for the low, moderate and high CAD prevalence populations, respectively. The segmental analysis yielded results with a similar trend.
The "results suggest that [CAA software] can be used in clinical practice to facilitate the accurate detection and exclusion of CAD on 64-slice and over CT scanners," said the study’s senior author Michael Poon, MD, director of advanced cardiac imaging in the department of radiology at Stony Brook. CAA software results "do not replace the expert analysis by the interpreting cardiologist, but rather work synergistically by combining the cardiologist's expertise with the software's capability," he said.