Artificial intelligence applications in photon-counting detector CT for coronary artery disease imaging: a narrative review

Coronary artery disease (CAD) remains the leading cause of morbidity and mortality worldwide. Coronary computed tomography angiography (CCTA) is a first-line non-invasive imaging modality for CAD evaluation, but conventional energy-integrating detector CT (EID-CT) is often limited by blooming artifacts, noise, and partial-volume effects. Photon-counting detector CT (PCD-CT) provides superior spatial resolution, reduced blooming, and multi-energy spectral data, yet its ultra-high-resolution (UHR) datasets frequently suffer from increased noise and complex spectral information that require advanced processing. Artificial intelligence (AI) has emerged as a promising solution to fully exploit PCD-CT's advantages. This narrative review aims to synthesize the current evidence on AI applications in PCD-CT for CAD imaging, evaluate their technical performance and clinical potential, identify limitations, and propose future directions.

A comprehensive literature search was conducted in PubMed, Web of Science, Scopus, and Embase, supplemented by hand-searching of reference lists. Original English-language studies applying AI techniques to PCD-CT data for CAD imaging were included, and categorized using a 5-level translational framework.

Thirty single-center studies were identified and grouped into three main categories. In image quality improvement, denoising models achieved 60-96% noise reduction while preserving spatial and spectral fidelity. Radiomics studies (the largest group) identified specific texture features in myocardium, plaques, pericoronary, and epicardial adipose tissue associated with coronary artery calcium severity, hypercholesterolemia, aging, and CAD presence, although feature reproducibility was only moderate. Automated detection/quantification/segmentation tools demonstrated high reproducibility, improved stenosis quantification, substantial workflow benefits, and promising CT-fractional flow reserve (FFR) performance with strong intra-individual agreement to EID-CT.

AI has shown promising technical advancements when applied to PCD-CT for CAD imaging. However, the evidence base remains limited by small sample sizes, single-center and single-scanner designs, lack of external validation, and absence of outcome-driven studies. Future research should focus on multicenter prospective validation, standardized protocols, human-in-the-loop systems, and clinical impact assessment to support the safe and effective translation of AI-enhanced PCD-CT into routine clinical practice.