论文部分内容阅读
目的研究建立人工神经网络模型用于评估大剂量甲氨蝶呤(high-dose methotrexate,HDMTX)化疗后的骨髓抑制程度,促进个体化用药。方法收集180例急性淋巴细胞白血病患儿行HDMTX化疗的临床资料。将所有资料随机分成2组,训练组(n=150):以化疗后中性粒细胞总数(NEU)减少率为输出目标,采用遗传算法配合动量法训练后建立人工神经网络;测试组(n=30):用建立的人工神经网络预测测试组患儿的NEU减少率,通过计算平均预测误差(MPE)、权重残差(WRES)、平均绝对预测误差(MAE)、平均预测误差平方(MSE)和均方根预测误差(RMSE)来验证模型。结果人工神经网络的MPE为(-2.05±7.41)%,WRES为(23.20±29.74)%,MAE为(6.12±4.53)%,MSE为(57.26±64.46)(%)2,RMSE为7.57%,有76.67%的病例相对预测误差在±20%以内。人工神经网络预测的准确度及精密度均优于多元线性回归模型(逐步回归法)。结论本研究建立的人工神经网络预测性能较好,可用于预测HDMTX化疗后骨髓抑制程度以指导个体化用药。
Objective To establish an artificial neural network model for evaluating the degree of myelosuppression after high-dose methotrexate (HDMTX) chemotherapy and to promote personalized medicine. Methods The clinical data of 180 patients with acute lymphoblastic leukemia treated with HDMTX chemotherapy were collected. All the data were randomly divided into two groups: training group (n = 150): the output of the total neutrophil count after chemotherapy (NEU) as the output target, using genetic algorithm with momentum training to establish artificial neural network; test group = 30): The NEU reduction rate of children in the test group was predicted by the established artificial neural network. The average prediction error (MPE), the weighted residuals (WRES), the mean absolute prediction errors (MAE), the average prediction error squares ) And root mean square prediction error (RMSE) to validate the model. Results The MPE of artificial neural network was (-2.05 ± 7.41)%, (23.20 ± 29.74)% for WRES, (6.12 ± 4.53)% for MAE, (57.26 ± 64.46)% for MSE, 7.57% for RMSE, There are 76.67% of cases relative prediction error within ± 20%. Artificial neural network prediction accuracy and precision are better than multiple linear regression model (stepwise regression). Conclusion The artificial neural network established in this study has good predictive performance and can be used to predict the degree of myelosuppression after HDMTX chemotherapy to guide individualized medication.