AI-CCTA for Predicting Cardiovascular Events in CAD

Background

Coronary artery disease (CAD) remains the leading cause of global mortality. While coronary computed tomography angiography (CCTA) is a Class I-recommended noninvasive modality for evaluating stable chest pain, its interpretation is often limited by inter-observer variability and physician experience. Although artificial intelligence (AI) assists in automating stenosis quantification and plaque analysis, current evidence lacks large-scale, real-world comparisons between different AI platforms using invasive coronary angiography (ICA) as the gold standard. Additionally, the interplay between CCTA-derived anatomical/functional parameters, systemic inflammation, and pharmacological modifications (e.g., statins, antiplatelet agents) in predicting long-term prognosis remains unclear.

Objectives

This study seeks to: (1) Perform a head-to-head comparison of different commercial AI software in diagnosing obstructive CAD using ICA as the reference; (2) Explore the correlation between AI-derived CCTA parameters (plaque characteristics, CT-FFR, pericoronary fat inflammation) and major adverse cardiac events (MACE); (3) Assess the modifying effects of lipid-lowering intensity, LDL-C attainment, antiplatelet regimens, and inflammatory markers on the prognostic value of CCTA.

Methods

This is a multicenter, retrospective cohort study involving 3,000 patients across five centers (Nanjing First Hospital, The Second Hospital of Jilin University, etc.). Patients aged ≥18 years who underwent CCTA and ICA within 90 days between June 2013 and June 2025 will be screened. CCTA images will be analyzed using commercially available AI platforms (Shukun, United Imaging) to extract parameters including stenosis grading, plaque volume, CT-FFR, and pericoronary fat attenuation index (FAI). Clinical data, laboratory results (lipids, hs-CRP, NT-proBNP), and medication histories will be retrospectively collected. The primary endpoint is MACE (a composite of cardiac death, myocardial infarction, unplanned revascularization, and hospitalization for unstable angina/heart failure). Follow-up duration will be at least one year.

Analysis

Sensitivity, specificity, and AUC will be calculated for AI-CCTA against ICA. Survival analysis (Kaplan-Meier and Cox regression) will be used to identify independent predictors of MACE. Interaction terms will be employed to evaluate the modifying effects of pharmacotherapies. Statistical analysis will be performed by the Department of Biostatistics, Nanjing Medical University.

Significance

By integrating anatomical, functional, and inflammatory dimensions of CCTA with real-world therapeutic data, this study aims to establish a comprehensive risk stratification framework for CAD, potentially guiding precision medicine approaches in clinical practice.

Inclusion criteria

• Age ≥18 years at the time of index coronary CT angiography (CCTA).
• Clinically suspected or confirmed coronary artery disease (CAD).
• Underwent both CCTA and invasive coronary angiography (ICA) within 90 days.
• Availability of clinical follow-up data for at least 12 months after the index CCTA.
• Sufficient image quality of CCTA and ICA to allow AI-based analysis and quantitative assessment.
• Signed informed consent or waiver of informed consent approved by the Institutional Review Board (IRB).

Exclusion criteria

• Age <18 years.
• Poor CCTA image quality precluding reliable AI analysis (e.g., severe motion artifacts, inadequate contrast opacification).
• History of prior coronary artery bypass grafting (CABG).
• Follow-up duration <12 months or incomplete follow-up records.
• Known hypersensitivity to iodinated contrast media, hyperthyroidism, severe hepatic or renal dysfunction (eGFR <30 mL/min/1.73m²), or malignancy.
• Pregnancy or lactation at the time of CCTA.
• Incomplete or missing core baseline data (e.g., lack of CCTA/ICA images, key clinical variables, or lipid profiles).