Objective To explore the application value of artificial neural network in the prediction of fetal distress. Methods A total of 198 maternal cases who were diagnosed as fetal distress before delivery, then received obstetric intervention in our hospital, and the neonatal Apgar score of whom was below 8 points were recruited, and all samples were randomly divided into two groups, with 99 cases in the training group and 99 cases in the validation group. Eleven input parameters were selected, namely, the character of amniotic fluid:clear, first degree, second degree and third degree. Fetal heart monitoring:normal, baseline variability weakened or disappeared, baseline >160 beats/min, baseline <110 beats per minute, early deceleration, deceleration variation, and advanced deceleration. The training group was constructed by numerical method, and then the validation group was used to test the diagnostic accuracy. Meanwhile, another 220 cases with prenatal diagnosis of fetal distress were as clinical false positive group, who were admitted to the hospital at the same period and received the obstetric intervention before delivery while after delivery did not show the intrauterine fetal distress, then the parameters were input to test the false positive rate. Results The diagnostic accordance rate was 87.88%, the false negative rate was 12.12%, the false positive rate was only 20.53%, and the diagnosis coincidence rate was 79.47%. Conclusion The application of artificial neural network to the prediction of fetal distress has research value and application prospects; increasing the input of clinical detection parameters and sample size may further improve the accuracy of the network. |