首页|期刊导航|上海中医药杂志|基于人工智能融合中医舌象、证素及西医指标构建2型糖尿病缓解预测模型

基于人工智能融合中医舌象、证素及西医指标构建2型糖尿病缓解预测模型OA

Construction of a predictive model for type 2 diabetes remission based on artificial intelligence integration of traditional Chinese medicine tongue manifestations,syndrome elements and Western medical indicators

中文摘要英文摘要

目的 旨在融合中医舌象参数、证素与西医指标,构建适用于营养干预实现2型糖尿病缓解的预测模型.方法 回顾性纳入2022年至2024年于上海中医药大学附属市中医医院内分泌科就诊且实施以糖尿病缓解为诊疗目的的营养干预治疗方案患者424例,收集基线及随访期间患者临床资料及舌象图片,构建四类特征组合(西医指标、西医指标+中医舌象参数、西医指标+中医证素、西医指标+中医舌象参数+中医证素),分别采用逻辑回归、随机森林和极限梯度提升算法进行建模,并通过受试者工作特征(ROC)曲线下面积(AUC)等指标评估模型性能.结果 采用随机森林算法构建的融合模型(西医指标+中医舌象参数+中医证素)预测性能最优,AUC可达0.95(95%CI:0.92~0.98),显著优于单一西医指标模型(AUC为0.89).变量重要性分析发现,空腹C肽、体重指数、定量胰岛素敏感性(QUICKI)、体脂率、糖化血红蛋白、腹部脂肪率以及中医舌象参数舌苔润泽度(taiL)、舌苔颜色参数(taiG)为较重要的预测变量.结论 研究成功构建了融合中医舌象参数、证素与西医指标的机器学习预测模型,该模型能识别适用于通过营养干预实现糖尿病缓解的优势人群.

Objective To construct a predictive model for the remission of type 2 diabetes through nutritional intervention by integrating traditional Chinese medicine(TCM)tongue manifestation parameters,syndrome elements,and Western medical indicators.Methods A retrospective analysis was conducted on 424 patients who underwent nutritional intervention targeting diabetes remission at the Department of Endocrinology,Shanghai Municipal Hospital of TCM Affiliated to Shanghai University of TCM from 2022 to 2024.Clinical data and tongue images of the patients at baseline and during follow-up were collected.Four types of feature combinations were constructed:Western medical indicators alone,Western medical indicators plus TCM tongue manifestation parameters,Western medical indicators plus TCM syndrome elements,and Western medical indicators plus TCM tongue manifestation parameters plus TCM syndrome elements.Logistic regression,random forest and extreme gradient boosting algorithms were respectively used for model construction,and model performance was evaluated using indicators including the area under the receiver operating characteristic(ROC)curve(AUC).Results The integrated model(Western medical indicators plus TCM tongue manifestation parameters plus TCM syndrome elements)constructed by the random forest algorithm exhibited the optimal predictive performance,with an AUC of 0.95(95%CI:0.92-0.98),which was significantly superior to the model with Western medical indicators alone(AUC=0.89).Variable importance analysis identified fasting C-peptide,body mass index(BMI),quantitative insulin sensitivity check index(QUICKI),body fat percentage,glycated hemoglobin(HbA1c),abdominal fat percentage,as well as TCM tongue manifestation parameters including tongue coating moistness(taiL)and tongue coating color parameter(taiG)as the most important predictive variables.Conclusion This study successfully constructed a machine learning predictive model integrating TCM tongue manifestation parameters,syndrome elements and Western medical indicators,which can identify the advantageous population suitable for achieving diabetes remission through nutritional intervention.

黄怡文;张珂;张倩为;曾先昌;朱烨琳;刘珍秀;江涛;陶枫

上海中医药大学附属市中医医院内分泌科(上海 200071)上海中医药大学附属市中医医院内分泌科(上海 200071)上海市嘉定区中心医院内分泌科(上海 201800)上海中医药大学附属市中医医院内分泌科(上海 200071)上海中医药大学附属市中医医院内分泌科(上海 200071)上海中医药大学附属市中医医院内分泌科(上海 200071)上海中医药大学中医学院(上海 201203)上海中医药大学附属市中医医院内分泌科(上海 200071)

糖尿病缓解人工智能机器学习预测模型舌诊中西医结合营养干预

diabetes remissionartificial intelligencemachine learningpredictive modeltongue diagnosisintegrated traditional Chinese and Western medicinenutritional intervention

《上海中医药杂志》 2026 (4)

1-9,9

上海市卫健委中医药传承创新发展三年行动计划项目[ZY(2025-2027)-1-2-4]上海市科委医学创新项目(23Y11921500)国家中医药管理局中医优势专科建设项目-老年病科(RCYS40192025002-Z)

10.16305/j.1007-1334.2026.z20251022001

评论