Automated quantification of mitral regurgitation by classifier design for contrast 2D echocardiography

The authors describe a classifier design and its performance validation for quantitative assessment of mitral regurgitation (MR) in intracavity contrast two-dimensional echocardiograms, based on microbubble motion analysis. A hierarchical linear classifier was built for four class MR classification using a training set of 202 contrast echocardiographic image sequences and some features corresponding to the changes of the microbubble quantity and distribution. After the classifier's performance was evaluated in test sets with a rotation method, the authors obtained the less biased performance estimates of respectively 61.8% and 90.5% for correct classification and partially correct classification, which show the applicability and potential usefulness of automated MR assessment.<
>