论文部分内容阅读
目的对胎儿超声参数进行神经网络处理,建立预测出生时的胎儿体重的前馈神经网络(BP)模型。方法本网络由输入层、隐藏层和输出层三部分组成,输入层有9个神经元(Neurons,Nij表示第i层第j个神经元),分别表示BPD、AD和FL的10th%、50th%和90th%。隐藏层分两层,第一层有18个神经元,第二层有6个神经元,这两层对输入信号进行处理。输出层有3个神经元,分别代表胎儿体重的10th%、50th%和90th%。用21个典型的信号模式训练网络,要求最大误差小于005,用Sigmoid公式转换BP的输入和输出信号。BP网络建成后用371例临床资料评价该BP模型。结果BP预测值与实际结果之间有明显的正相关(r=06850,P<0001),预测误差平均为318g,约1097%。误差小于100g的占212%,小于200g的占3908%,小于300g的占5346%,小于500g的占7926%。BP的准确性受胎儿出生体重影响(r=06,P<005),但与胎龄、超声检查与分娩的时间差等因素无关。结论BP网络预测胎儿体重十分有效,但网络的结果以及参数的设定等方面有待于进一步的改进。本文为超声参数分?
OBJECTIVE: To carry out neural network processing of fetal ultrasound parameters and establish a feedforward neural network (BP) model for predicting the birth weight of the fetus. Methods The network consists of three parts: input layer, hidden layer and output layer. There are 9 neurons in the input layer (Neurons, Nij denotes the jth neuron in the ith layer), representing the 10th%, 50th % And 90th%. Hidden layer is divided into two layers, the first layer has 18 neurons, the second layer has 6 neurons, the two layers of the input signal processing. The output layer has 3 neurons, representing the 10th, 50th and 90th% of the fetal weight, respectively. Training the network with 21 typical signal patterns requires a maximum error of less than 005 and converts the BP input and output signals using Sigmoid’s formula. BP network built with 371 cases of clinical data to evaluate the BP model. Results There was a significant positive correlation between BP prediction and actual results (r = 0.6850, P <0.001). The average prediction error was 318 g (10.97%). Errors less than 100g accounted for 21 2%, less than 200g accounted for 39 08%, less than 300g accounted for 53 46%, less than 500g accounted for 79 26%. The accuracy of BP was affected by fetal birth weight (r = 0.6, P <005), but not with gestational age, time of sonographic examination and poor delivery. Conclusion BP neural network prediction of fetal weight is very effective, but the results of the network and the parameters set needs to be further improved. This article for the ultrasonic parameters points?