Functional analysis of mitral complex geometry using Support Vector Machines from 3D echocardiography

In order to assist diagnosis and surgical repair of congenital mitral disease, quantitative analysis of 3D geometry of the mitral complex is necessary for better understanding mechanism and dysfunction of the mitral complex. This work aims to extract geometric parameters of mitral complex and utilize Support Vector Machines (SVM) based classifier to support diagnosis of congenital mitral regurgitation (MR). With a control group of 20 normal young children (11 boys, 9 girls, 5.96±3.12 years) with normal structure of mitral apparatus, 20 patients (9 boys, 11 girls, 5.59±3.30 years) suffering from severe congenital MR are recruited in this study. The results of parameter validation demonstrates that the measurement precision is in the range of inter-/intra-observer variability. SVM-based classifier achieves average classification accuracy at 85.0% in the present population.