408 Stress Cardiac Magnetic Resonance With In-Line AI Quantitative Perfusion Mapping; Clinical Experience for Ischaemia and Microvascular Dyfunction

Abstract

Adenosine Stress Perfusion Cardiac Magnetic Resonance (CMR) is a highly accurate technique for non-invasive assessment of myocardial perfusion, as shown in recent CEMARC, MR-INFORM, SPINS and GADACAD trials. Quantitative perfusion mapping using convoluted neural network artificial intelligence (AI) reconstruction has been proposed to improve accuracy and diagnostic performance. 33 consecutive stress scans CMR clinically referred for evaluation of angina pectoris, performed on a 1.5T Siemens Aera system. Standard sequence was a free-breathing motion corrected FLASH dynamic perfusion after 140mcg/kg adenosine, compared to an investigational work-in-progress myocardial blood flow (MBF) dual bolus technique with automatic arterial input function quantitated in ml/min/g using 16-segment AHA model. 22 scans referred for ischaemia, 11 for microvascular dysfunction. Level-3 expert reader blindly evaluated grey-scale dynamic perfusion imaging, and separately evaluated quantitative perfusion maps. Scans were categorised as ischaemia, microvascular, or normal coronary physiology. Diagnostic confidence was measured on a Likert rating scale from 1 (unconfident) to 3 (very confident). 31/33 scans (94%) had successful perfusion mapping, with 2 excluded due to artefact from arrythmia or failed motion correction. 20/22 scans referred for ischaemia (91%) mapping improved diagnostic confidence (score 2.03 vs 2.88, P<0.001). For all 11 (100%) microvascular dysfunction cases perfusion mapping was diagnostic over standard imaging. Quantitative perfusion mapping improves diagnostic confidence and allows accurate separation of ischaemic vs microvascular causes of angina.