Optimal cut-off point of artificial intelligence-based global circumferential strain measured at stress for prediction of cardiovascular events

Abstract Background Global circumferential strain (stress-GCS) measured during vasodilator stress cardiovascular magnetic resonance (CMR) has shown incremental prognostic value to predict major adverse cardiovascular events (MACE) above traditional stress CMR factors. However, no prior study aimed to define the optimal cut-off of stress-GCS. Purpose To determine the optimal cut-off point of stress-GCS to predict MACE in a large cohort of patients with normal stress CMR. Methods Between 2017 and 2018, all consecutive patients with normal stress CMR defined by the absence of inducible ischemia and late gadolinium enhancement (LGE) were included retrospectively. Stress-GCS was measured using an artificial intelligence (AI)-fully automated machine-learning algorithm based on featured-tracking imaging from short-axis cine images. The primary composite outcome was MACE defined by cardiovascular mortality or nonfatal myocardial infarction (MI). A survival tree method was used to identify the optimal cut-off for stress-GCS. Results In 1,335 patients (65±12 years, 67% male), 52 (3.9%) experienced a MACE after a median follow-up of 5.1 (4.8-5.4) years. The best cut-off of stress-GCS to predict MACE was -10%. After adjustment for traditional risk factors and stress left ventricular ejection fraction (LVEF), stress-GCS≥-10% was independently associated with MACE (adjusted HR, 12.4 [95% CI, 5.89-26.1], p