融合舌脉诊客观参数的胃癌前病变高危人群预测模型构建研究  

Construction of a predictive model for high-risk populations of precancerous lesion of gastric cancer based on characteristic integrated with traditional Chinese medicine tongue and pulse diagnosis parameters

作　　者：肖雯迪 于莉 叶金连[1,2,4] 杨小婷[1,2] 肖逸菲 王洋[1,2] 林雪娟[1,2] 李灿东[1,2] XIAO Wendi;YU Li;YE Jinlian;YANG Xiaoting;XIAO Yifei;WANG Yang;LIN Xuejuan;LI Candong(Research Base of Traditional Chinese Medicine Syndrome,Fujian University of Traditional Chinese Medicine,Fuzhou 350122,China;Key Laboratory on TCM Health Differentiation in Fujian Province(Fujian University of Traditional Chinese Medicine),Fuzhou 350122,China;The Third Affiliated People’s Hospital of Fujian University of Traditional Chinese Medicine,Fuzhou 350108,China;The Second Affiliated Hospital of Fujian University of Traditional Chinese Medicine,Fuzhou 350003,China;Integrated Rehabilitation Group,Seattle 98107,USA)

机构地区：[1]福建中医药大学中医证研究基地,福州350122 [2]福建省中医健康状态辨识重点实验室(福建中医药大学),福州350122 [3]福建中医药大学附属第三人民医院,福州350108 [4]福建中医药大学附属第二人民医院,福州350003 [5]综合康复集团,华盛顿州西雅图98107

出　　处：《中华中医药杂志》2025年第1期170-179,共10页China Journal of Traditional Chinese Medicine and Pharmacy

基　　金：国家自然科学基金联合基金项目(No.U1705286);国家中医药管理局第二届全国名中医传承工作室建设项目(No.国中医药办人教函[2022]245号)。

摘　　要：目的:通过套索算法(LASSO)联合Logistic回归的方法筛选胃癌前病变(PLGC)危险因素,建立融入舌脉客观参数的PLGC风险预测可视化模型。方法:纳入慢性胃炎患者319例,收集患者的基础信息、疾病资料、舌脉象客观参数并由临床四诊信息提取证素辨识结果。根据胃黏膜病理结果将患者分为PLGC组146例和非PLGC组173例,比较两组患者在上述资料间的差异。使用LASSO回归方法筛选危险因素,并在此基础上通过多因素Logistic回归方法筛选PLGC高危因素,最后根据PLGC高危因素构建列线图模型并进行验证与评价。结果:两组患者在基础信息、疾病资料、舌脉象客观参数及证素方面通过单因素分析筛选出36个危险因素(P<0.05,P<0.01);LASSO回归联合多因素Logistic回归分析再次筛选,显示PGⅠ、舌色淡、舌苔a值、舌苔b值、t5、w/t、证素肾和血瘀为发生PLGC高危因素(P<0.01)。基于上述高危因素构建PLGC高危人群预测列线图模型,模型效能评估显示曲线下面积AUC为0.872(95%CI为0.833~0.912),Bootstrap验证模型预测概率和实际观测概率基本吻合。结论:使用PGⅠ、舌色淡、舌苔a值、舌苔b值、t5、w/t、证素肾和血瘀为高危因素建立的PLGC高危人群预测模型具有较好的区分度和校准度,模型融入舌脉诊客观参数,有助于临床医生动态监测慢性胃炎患者胃黏膜病变情况并及时施治。Objective:The risk factors of precancerous lesion of gastric cancer(PLGC) were screened by LASSO algorithm and Logistic regression,and the visual model of PLGC risk prediction was established by incorporating objective parameters of tongue and pulse.Methods:A total of 319 patients with chronic gastritis were included in this study.The basic information,disease data,objective parameters of tongue and pulse condition were collected.And according to the patients' clinical four diagnostic information extraction syndrome element identification results.The patients were divided into PLGC group(146 cases) and non-PLGC group(173 cases) according to the pathological results of gastric mucosa.The risk factors of PLGC were screened by LASSO regression,and the high risk factors of PLGC were screened by multivariate Logistic regression.Finally,a nomogram model was constructed and validated.Results:Thirty-six risk factors were screened out by univariate analysis in basic information(P<0.05,P<0.01),disease data,objective parameters of tongue and pulse condition and syndrome elements,and again by LASSO regression combined with multivariate Logistic regression analysis,the results showed that PG I,pale tongue color,a value of tongue coating,b value of tongue coating,t5,w/t,kidney of syndrome element and blood stasis of syndrome element were the high risk factors of PLGC(P<0.01).The predictive nomogram model of PLGC high risk population was constructed based on the above risk factors.The AUC of area under the curve was 0.872(95%CI 0.833-0.912),the prediction probability of Bootstrap validation model is basically consistent with the observed probability.Conclusion:The prediction model of PLGC high risk population was established by using PG I,pale tongue color,a value of tongue coating,b value of tongue coating,t5,w/t,syndrome element kidney and blood stasis as high risk factors.The model is integrated into the objective parameters of tongue and pulse examination,which is helpful for the clinician to monitor the pathological change

关 键 词：风险预测模型 胃癌前病变 舌诊客观化特征 脉诊客观化特征 列线图 

分 类 号：R273[医药卫生—中西医结合]