Application of artificial neural network for estimation of fetal weight

Accurate estimation of fetal weight is crucial for deciding which mode is the best for babies to be delivered. With the advancement of ultrasonic technologies, sonographic parameters of the fetal can be used to estimate fetal weight. Fetal weight estimated by regression methods is relatively acceptable in the clinical Obstetrics, but the accuracy of estimated fetal weight remains to be improved. This study was aimed to develop a group-based artificial neural network model to improve the accuracy of fetal weight estimation through sonographic parameters. Stepwise regression analysis was used to examine and extract the significant parameters. The input layer in the artificial neural network model included seven significant parameters such as biparietal diameter, occipito-frontal diameter, abdominal circumference, gestational age, femur length, gender, and fetal presentation. A total of 2,107 consecutive singleton fetuses were divided into training group with 1,411 samples and testing group with 696 samples. The results show that the accuracy of fetal weight estimated by the artificial neural network model is significantly better than those by regression methods. The importance of this study is to consider and control the heterogeneity among the high variability and broad ranged parameters by statistics, and to choose scientific parameters as reasonable input variables of artificial neural network to improve the estimation of fetal weight. This study has proved the accuracy of fetal weight estimation by artificial neural network model is better than those of previous models.