摘要
目的通过评估全数字化分析中医基本症候的准确率,来验证全数字化分析中医基本证候与传统中医理论对于多囊卵巢综合征(polycystic ovary syndrome,PCOS)的分型有较高的一致性,从而可以客观地辅助临床对PCOS的诊疗。方法收集PCOS病例,数字化记录症状、体征、辨证分型结果;并采用多标记学习法中的多标记k近邻方法(multi-label k-nearest neighbor,ML-kNN)、多标记贝叶斯学习算法(multi-label naive Bayesian,MLNB)法和深度森林算法(multi-grained cascade forest,gcForest),通过计算平均准确率(average precision)、覆盖距离(coverage)、汉明损失(Hamming loss)、首标记错误和排序损失(ranking loss)5个数值来评估全数字化分析中医基本症候的准确率。结果用ML-kNN、MLNB和gcForest将临床采集的数据建立数学模型,经过计算后得出158例PCOS确诊患者中医临床辨证分型为肾虚、脾虚、肝郁、痰湿和血瘀,其中肾虚无兼症的患者52例,肾虚和肝郁并存的患者48例,肾虚和痰湿并存的患者58例。用ML-kNN得出的证型准确率分别为:肾虚66.6%±10.2%、脾虚86.15%±2.9%、肝郁59.8%±9.7%、痰湿72.2%±11.6%,血瘀82.4%±4.6%。用MLNB得出的证型准确率分别为:肾虚65.5%±8.0%、脾虚85.6%±7.1%、肝郁74.2%±7.7%、痰湿70.5%±4.5%,血瘀81.8%±7.7%。用gcForest得出的证型准确率分别为:肾虚87.2%±5.0%、脾虚86.6%±4.8%、肝郁79.2%±6.5%、痰湿79.4%±6.8%,血瘀82.3%±5.9%。结论用中医信息系统计算的PCOS的中医症候有肾虚、脾虚、肝郁、痰湿、血瘀,与曹玲仙教授对PCOS的分型有较高的一致性。说明全数字化采集PCOS患者证候信息并通过现代数据挖掘方法进行辨证论治,可以对PCOS中医临床证候进行有效规律总结,对临床诊疗有一定的帮助。
Objective By evaluating the accuracy of full digital analysis of basic symptoms of traditional Chinese medicine,to verify that the full digital analysis of basic symptoms of traditional Chinese medicine is basically consistent with the classification of polycystic ovary syndrome(PCOS)in traditional Chinese medicine theory,so as to objectively assist the clinical diagnosis and treatment of PCOS.Methods PCOS cases were collected and the symptoms,signs and syndrome differentiation results were recorded digitally.The multi labeled k-nearest neighbor method(ML-kNN),multi labeled Bayesian learning algorithm(MLNB)and multi-grained cascade forest algorithm(gcForest)were used to evaluate the accuracy of fully digital analysis of basic symptoms of traditional Chinese medicine by calculating five values:average precision,coverage,Hamming loss,first labeling error and ranking loss.Results ML-kNN,MLNB and gcForest were used to establish a mathematical model based on the clinical data collected.After calculation,158 patients with PCOS were divided into kidney deficiency,spleen deficiency,liver depression,phlegm dampness and blood stasis,including 52 patients with no concurrent disease of kidney deficiency,48 patients with kidney deficiency and liver depression,and 58 patients with kidney deficiency and phlegm dampness.The accuracy rates of syndrome types obtained by ML-kNN were:kidney deficiency 66.6%±10.2%,spleen deficiency 86.15%±2.9%,liver depression 59.8%±9.7%,phlegm dampness 72.2%±11.6%,blood stasis 82.4%±4.6%.The accuracy rates of syndrome types obtained by MLNB were:kidney deficiency 65.5%±8.0%,spleen deficiency 85.6%±7.1%,liver depression 74.2%±7.7%,phlegm dampness 70.5%±4.5%,blood stasis 81.8%±7.7%.The accuracy rates of syndrome types obtained by gcForest were:kidney deficiency 87.2%±5.0%,spleen deficiency 86.6%±4.8%,liver depression 79.2%±6.5%,phlegm dampness 79.4%±6.8%,blood stasis 82.3%±5.9%.Conclusions The TCM symptoms of PCOS calculated by TCM information system include kidney deficiency,spleen def
作者
郭姗珊
陈厚儒
王书云
姚笛
吴胜男
朱光耀
俞而慨
颜建军
GUO Shanshan;CHEN Houru;WANG Shuyun;YAO Di;WU Shengnan;ZHU Guangyao;YU Erkai;YAN Jianjun(Longhua Hospital Shanghai University of Traditional Chinese Medicine,Shanghai 200032;East China University of of Science and Technology,Shanghai 200237;First Maternity and Infant Hospital,Affiliated to Tongji University,Shanghai 200092)
出处
《北京生物医学工程》
2023年第2期170-177,共8页
Beijing Biomedical Engineering
基金
上海市科技计划项目(17401932000)资助。
关键词
多囊卵巢综合征
中医信息系统
中医分型
多标记学习法
polycystic ovary syndrome
TCM information system
TCM classification
multi-marker learning method
