| Abstract | This study was conducted to determine if single and multiple ultrasonographic examinations in conjunction with a neural network (NN) are useful for predicting IUGR. To demonstrate applicability, symmetric IUGR and asymmetric IUGR curves are used for the input of the NN. Nominal curves of HC, AC, and HC/AC ratio vs. gestational age for normal, SGA, and LGA, and transition regions for symmetric IUGR and asymmetric IUGR are coded on the connection weights of an NN. Weekly examinations (from 1 to 4 weeks) are considered. An increasing success rate for the increasing number of examinations (weekly) are obtained. |