Poster R16 - Quantitative comparison of precision of echocardiography versus cardiac magnetic resonance imaging in border-length-based global longitudinal strain and left ventricular ejection fraction

Introduction: There are many scenarios where early detection of changes in cardiac function is critical for patient care and survivorship, but this is limited in part by inherent measurement uncertainties that vary with the type of imaging modality, functional metric and reader experience Two key questions that heretofore have not been adequately addressed are: first, what is the minimum change in left ventricular (LV) ejection fraction (LVEF) and border length global longitudinal strain (GLSBL) that can be reliably detected with cardiac magnetic resonance imaging (cMRI) and echocardiography (EC)? Second, what are the conditions under which one metric-modality pair is better than another for detecting subtle LV dysfunction?

Materials and

Methods: Representative cMRI and EC image sets were acquired in a single human subject. Six experienced and three inexperienced readers manually segmented the endocardial border in the two-chamber and four-chamber views for each modality. Inter-reader variability in single-plane and biplane estimates for GLSBL and LVEF were computed for both EC and cMRI. An in-silico heart model was also used, to interrogate from first principles the relationship between border uncertainty and metric precision.

Results, Conclusions, and Discussions: Comparing modalities, for biplane estimates of GLSBL the inter-reader variability for experienced readers was 37% lower with cMRI versus EC: 6.0% versus 9.5%. For biplane estimates of LVEF, cMRI was associated with a 75% reduction in variability versus EC (3.4% versus 13.4%). Comparing metrics, with cMRI, biplane estimates of LVEF were 43% less variable (more precise) than GLSBL (3.4% versus 6.0%). But with EC, LVEF was 22% more variable (less precise) than GLSBL (13.4% versus 9.5%). GLSBL is the more appropriate metric for EC, while LVEF is the more appropriate metric for cMRI. The combination of cMRI with LVEF gave a 2.8-fold improvement in detection sensitivity over EC with GLSBL (3.4% versus 9.5%) and is therefore recommended for assessing small changes in heart function. The EC-based LVEF and GLS results concur with prior clinical findings. The results inform imaging and assessment approaches for optimized patient care and guide the design of future clinical studies.

Acknowledgements (Optional): Research was provided from grants through the Florida Breast Cancer Foundation, the Ocala Royal Dames for Cancer Research, the University of Florida Clinical Translational Science Institute’s Human Imaging Core, and the University of Florida University Scholars Program.