The decision of intrauterine growth retardation from ultrasonographic examinations with neural networks

Abstract

Our purpose is to make decision of intrauterke growth retardation (IUGR) through siagle and multiple uftrasonographic fetal growth assessments using a neural network (NN). This study was undertaken to show if a feedforward NN can leam nominal growth curves of head circumference (HC), abdominal circumference (AC), and HC/AC ratio versus gestatioaal age and can help doctors in diagnosis of IUGR. Weekly (from 1 to 4 weeks) ultrasonographic examinations are taken as input to NN. A feedforward NN is used as a function approximates. Back propagation (BP) algorithm is used to optimize connection weights using samples from nominal curves. It was observed that a NN can improve the accuracy of the decision of IUGR by the multiple weekly examinations which mean monitoring the dynamic process of a change in size over time. ît was concluded that the applicability of NNs to determination of IUGR is possible and it is a fruitful line of inquiry for further work.