Accuracy and consistency of subtraction dual-layer spectral computed tomography in diagnosing coronary stenosis: invasive coronary angiography validation study

Background: Noninvasive assessment of coronary artery stenosis is critical for the diagnosis and management of coronary artery disease (CAD), particularly in patients with intermediate pretest probability for CAD or contraindications to invasive procedures. However, conventional coronary computed tomography angiography (CCTAcon) is limited by the presence of artifacts from calcified plaques, leading to overestimation of stenosis severity. Subtraction CCTA (CCTAsub) has the potential to overcome these limitations by eliminating these artifacts and improving the visualization of the vessel wall, thereby enhancing diagnostic accuracy in terms of percent diameter stenosis (DS). The objective of our study was to evaluate the feasibility of using CCTAsub with dual-layer spectral CT to improve coronary vessel wall visualization and the diagnostic accuracy across different plaque types. Methods: Seventy-one consecutive patients with suspected or known CAD who underwent both CCTA and invasive coronary angiography (ICA) within 30 days participated in this study. Subtraction images were generated via the subtraction of virtual noncontrast datasets from 70-keV datasets of dual-layer spectral computed tomography (CT). Two independent radiologists evaluated the image quality and the conspicuity of the inner and outer vessel walls in subtraction images. ICA served as the gold standard for DS assessment. Interobserver agreement for subjective image quality was assessed using weighted kappa statistics. The diagnostic accuracy of CCTAsub in measuring DS was evaluated via receiver operating characteristic curve analyses. Results: The study population comprised 45 males (mean age: 64.5±11.5 years) and 26 females (mean age: 67.8±9.6 years), with 62.1% having severe DS (≥50%) as confirmed by ICA. The subjective evaluation of subtraction images yielded high scores for image quality (2.8±1.8), inner vessel wall conspicuity (1.8±0.8), and outer vessel wall conspicuity (1.9±0.9). Radiologists’ subjective scores showed good consistency (all kappa values ≥0.7). Compared to CCTAcon, CCTAsub demonstrated higher agreement in measuring the DS of coronary arteries (intraclass correlation coefficient: 0.96 vs. 0.60). CCTAsub also demonstrated robust performance in accurately detecting DS [area under the curve: 0.986; 95% confidence interval (CI): 0.955–0.998], with a sensitivity of 98.2% and a specificity of 87.9%. Conclusions: Subtraction images derived from dual-layer spectral CT improved the visualization of the coronary artery vessel wall and showed good agreement with ICA in assessing the DS of coronary arteries. The findings support the clinical utility of CCTAsub for accurate stenosis quantification, potentially reducing the need for invasive diagnostics in select patients.