Classifying tissue and structure in echocardiograms

The authors present a system to automatically classify structures and tissues in echocardiogram images. The approach uses a multiple-feature, hierarchical, fuzzy, neural network fusion solution to the problem. The technique appears to work well. The technique is general and can be applied to a variety of image processing problems not only in the medicine but other areas as well. The processing presented here is computationally expensive; every pixel is processed independently. However, there are several ways to improve performance. The most obvious is to take advantage of the parallelism. In addition, every pixel need not be processed. If only large cardiac structures, such as the left ventricle, need be classified, then processing of the compressed image through the binary nets is sufficient. Neighboring class information may be used as well. Finally, the authors have not accounted for time in the processing. There is a high degree of redundancy in the successive frames of video images. It is anticipated that including temporal information will increase both statistical and computational performance. Temporal information will also force smoothing of the images over time and give better classification performance to even those areas that do not move.<
>