Fully automated assessment of left ventricular volumes, function and mass from cardiac MRI

The importance of quantification of left ventricular (LV) size, function and mass is increasingly recognized through growing evidence about the prognostic value of these indices and their diagnostic role in patient follow-up during therapy. However, quantitative evaluation from cardiac magnetic resonance (CMR) images relies on manual tracing of LV endo- and epicardial boundaries, which is subjective and time-consuming. Our goal was to develop a fully automated technique for the detection of these boundaries to assess LV volumes, ejection fraction (EF) and mass. Our automated approach consists of: 1) identification of the LV cavity based on detection of moving and circular structures in short-axis views; 2) endocardial detection using a region-based probabilistic level set model to allow volume measurements throughout the cardiac cycle; 3) epicardium detection at end-diastole based on an edge-based level set model to allow LV mass measurement. This approach was tested in 10 patients by comparing automatically derived LV volumes, EF and mass using manual tracing as a reference. Automated detection of the endo- and epicardial boundaries took <;5 minutes per patient on a standard PC. The detected boundaries were in good agreement with manual tracing. As a result, LV volumes, EF and mass showed good inter-technique concordance, reflected by minimal biases and narrow limits of agreement. The proposed technique allows fully automated, fast and accurate measurements of LV volumes, EF and mass from CMR images, which may address the growing clinical need for quantitative assessment.