疗效预测方法在针刺治疗抑郁症研究中的应用探讨  

Application of artificial neural network in forecasting effectiveness of acupuncture therapy for depression

作　　者：杨秀岩[1] 图娅[1] 段冬梅[2] 

机构地区：[1]北京中医药大学针灸学院,北京100029 [2]解放军总医院

出　　处：《北京中医药大学学报》2008年第5期355-357,共3页Journal of Beijing University of Traditional Chinese Medicine

基　　金：北京市自然科学基金重点项目(No.7051003)

摘　　要：目的探讨应用神经网络方法进行针刺治疗抑郁症的疗效预测。方法使用量表工具,应用神经网络方法进行疗效预测。选择与抑郁症针刺治疗效果密切相关的因素以及体现中医学整体思想的疾病状态、生活质量等指标作为神经网络的输入变量,将其量化数据赋予网络的输入层,将经治疗后量表的评定结果作为输出变量值。收集病例形成神经网络的训练和测试样本,进行网络的训练和测试,评估网络性能,讨论其临床应用性。结果网络训练误差达到预期目标(均方误差mse=0.0010),对HAMD减分率和SDS减分率的预测拟合度较好,且拟合精度较高;表明网络测试的拟合程度比训练结果的拟合程度差,网络模型对测试样本的预测误差大于对已知样本的预测误差。结论网络模型具有较好的学习能力,而对未知样本的识别能力不及对已知样本的识别能力。通过增加样本量、改进网络模型等途径,有望为临床提供可靠性良好的预测方法。objective To discuss the application of artificial neural network in forecasting effectiveness of acupuncture therapy for depression.Method A forecasting method based on neural network was used in an underway research which was aimed at predicting the effectiveness.The quantized values of relevant factors of effect,disease states and quality of life were chosen as input variables of the neural network,while the changes in results of the scales after treatment were set as output variables.The collected cases were divided into training samples and testing samples for training and testing the neural network.Results The training result of the network reached the expected goal（mse=0.001）,the recognition about reducing score rate of HAMD and SDS was of good curve fitting and high fitting precision;the testing result was not so good as the training result,the recognition about the unknown samples showed worse curve fitting and more forecast error than that about the known samples.Conclusion The neural network model in this research has got rather ideal results of estimation in identification for the known samples,however,its predicting ability for the unknown samples has not achieved such ideal results as that for known samples.The forecast model based on neural network is expected to be a helpful method for clinical research and practice with the work of expanding the sample size and improving the neural network model.

关 键 词：抑郁症 针刺 疗效 预测 神经网络 

分 类 号：R246.6[医药卫生—针灸推拿学]