Echocardiogram processing and classification using data-dependent systems analysis

This paper presents results of a study on M-mode echocardiogram analysis and classification using the recently developed data-dependent system (DDS) methodology along with clustering techniques. Results of thirty clinically obtained waveforms yielded a correct classification rate of between 83% and 90%, depending on the features used for clustering. The choice of features was facilitated, in part, by a study of six hypothetical waveforms representing an echocardiogram in transition from normal to stenosed. The DDS approach offers potential in the identification and recognition of various cardiac abnormalities, based on echocardiographic data. A comparison with Fourier analysis confirmed the results and points to the advantages of DDS in (a) minimizing false negatives and (b) providing additional features for improved discrimination.