Detecting the wall-motion of the fetal heart within ultrasound images

Abstract

An algorithm has been developed for automatically detecting the chamber and internal walls of the fetal heart from ultrasound images. This algorithm was a hybrid in that it had elements of both computational and knowledge-based image understanding. A combination of multiresolution edge-detection, active contours, pixel labelling and belief networks was employed. Use of these techniques greatly improved boundary detection over an exclusively contour based method without increasing computation time markedly. Following the use of these techniques the results of the boundary detection were much improved. It was found that measurements of heart-rate could be derived from the detected boundaries.