Quantitative assessment of mitral regurgitation by automated contrast 2D echocardiography

Abstract

Describes some approaches to the adaptive detection of microbubbles in contrast 2D echocardiograms and mathematical descriptions of microbubble characteristic movement patterns for quantitative assessment of mitral regurgitation (MR). A hierarchical linear classifier was designed by using features selected from the descriptions for distinguishing 4-class MR. With 202 image sequences, the classifier's performance was validated by using substitution and rotation methods yielding a sensitivity (SE) of 85.9%, a specificity (SP) of 66.8%, and an overall correct classification rate of 61.8%, as well as partially correct SE of 95%, SP of 89.8%, and overall classification rate of 90.5%, which are clinically acceptable.<>