| abstract |
Methods and apparatus for categorization of the diastolic dysfunction of a heart into a classification in a set of classifications are provided in which a plurality of ultrasound gray-scale measurement images of the heart across a plurality of heartbeats is obtained. A plurality of cardiac parameters is determined from the measurement images. In this determination, two or more of the images contributes to each cardiac parameter. The parameters are subjected to a linear discriminant function thereby obtaining a first prediction of the classification. The parameters are also subjected to a weighted neighborhood scheme thereby obtaining a second prediction of the classification. The parameters are further subjected to an artificial neural network thereby obtaining a third prediction of the classification. The first, second, and third prediction are applied to a majority voting method thereby obtaining the classification, in the set of classifications, for the diastolic dysfunction of the heart. |