Semi-automatic inline calculation of left ventricular function using cardiac MRI (CMR)

Background Cardiac Magnetic Resonance (CMR) imaging is increasingly recognized as the gold standard for evaluating left ventricular function (LVF). Measurement of LVF is typically done manually by tracing endocardial and epicardial borders on a stack of short axis (SA) cine images of the heart and is based on the modified Simpson assumption that the left ventricle approximates a spherical shape. A drawback of this method is that it is laborious and timeconsuming. The recent introduction of computer-based inline VF software that allows for automatic assessment raised the prospect for significant reductions in CMR post processing times. Initial results have found that fully automatic analysis is unreliable as compared to manual calculations. However, usually only 1 or 2 contours are inaccurate using the automatic software suggesting that, if these are corrected manually, a semi-automatic approach may be helpful. This study sought to compare fully automatic inline VF tracking to this semi-automatic method using the manual technique as gold standard.