Segmentation of ultrasound images by using an incremental self-organized map

Abstract

A new incremental self-organized map is proposed for the segmentation of the ultrasound images. Elements of the feature vectors are formed by the fast Fourier transform (FFT) of image intensities in 4/spl times/4 square blocks. In this study, two neural networks for segmentation are comparatively examined: Kohonen map, and incremental self-organized map (ISOM). It is observed that ISOM gives the best classification performance with less number of nodes after a short training time.