人工神经网络诊断帕金森病的应用研究  

Study on the diagnosis of Parkinson's disease with artificial neural network

作　　者：常崇旺[1] 高国栋[1] 陈洪[1] 李维新[1] 

机构地区：[1]解放军第四军医大学唐都医院神经外科,陕西省西安市710038

出　　处：《中国临床康复》2003年第28期3818-3819,共2页Chinese Journal of Clinical Rehabilitation

摘　　要：目的:探讨人工神经网络模拟医学专家诊断帕金森病的应用价值。方法:以人工神经网络作为诊断预测手段,将帕金森病的诊断成立与否作为实际输出,以误差反向传播网通过对样本的一般情况、危险因素、病史、症状、体征、检验结果、药物治疗情况7大类共59项特征指标进行拟合获得预测结果。结果:网络训练误差指标0.03;检测样本与专家诊断对照χ2=0.03,P>0.975;网络诊断误诊率6.4%,漏诊率8.3%,准确度92.9%。结论:人工神经网络能够模拟医学专家进行临床诊断,是一种很有前途的帕金森病诊断方法。AIM:To explore the value of artificial neural network as medical experts in diagnosis of Parkinson' s disease.METHODS:Artificial neural network was taken as the predicting diagnosing tool,the diagnosing result of Parki nson' s disease as expected output.The predicting results were obtained th rough fitting of 7kinds,including 59characteristic indices,such as personal information,risk factors,me dical records,symptom,character,resul ts of testing,effect of medication.RESULTS:The network training error index was0.03;the result of network diagnosis was not significantly dif ferent from that of expert diagnosis(χ 2 =0.03,P>0.975);the misdiagnosis rate of the network was 6.4%,rate of missed diagnosis was 8.3%,the accur acy was 92.9%.

关 键 词：人工神经网络 诊断 帕金森病 药物治疗 

分 类 号：R742.5[医药卫生—神经病学与精神病学]