Deep-learning reconstruction to improve image quality of myocardial dynamic CT perfusion: comparison with hybrid iterative reconstruction

Myocardial computed tomography (CT) perfusion (CTP) is a promising approach to detect myocardial ischaemia as a predictor of the functional relevance of coronary artery disease (CAD).1–3 There are currently two approaches in myocardial CTP imaging: static and dynamic. Static CTP, which evaluates contrast medium distribution in the left ventricle (LV) myocardium at a single time point, is easy to implement but limited in its ability to quantify myocardial blood flow (MBF).4 Dynamic CTP, on the other hand, allows quantification of MBF and has shown promise in accurately detecting haemodynamically significant CAD2,3; however, dynamic CTP techniques involve the acquisition of multiple consecutive phases and, therefore, the radiation dose is generally higher than that of static CTP techniques.