摘要
目的基于多种机器学习算法探究电子鼻(Electronic nose,E-nose)呼出气对结直肠癌(Colorectal cancer,CRC)早期检测的准确性,及对CRC常见中医证素的初步辨识能力。方法研究设计为前瞻性、观察性研究,共招募125名CRC患者和107例健康受试者。收集四诊信息,通过证素辨证方法分析中医证素分布情况,运用Cyranose 320便携式电子鼻获取患者口腔呼气气味图谱,结合随机森林、K最近邻、逻辑回归、支持向量机、极端梯度提升5种经典机器学习算法,对CRC发病及其常见病位、病性证素进行气味图谱辨识。结果(1)结直肠癌常见病位证素从高到低依次是大肠、脾、肝、胃,病性证素从高到低依次是气虚、阳虚、气滞、阴虚。(2)随机森林模型能有效辨识CRC与健康受试者,准确度为88.57%、特异度为80.56%、灵敏度为97.06%、AUC为0.89(95%CI:0.81-0.98)。(3)随机森林模型对CRC常见病位、病性证素的呼吸图谱辨识效能均最佳,分类准确度普遍达90.00%以上。结论基于E-nose呼气图谱不仅能实现对CRC的早期无创检测,且能为CRC中医证素诊断提供客观、数智化研究证据。
Objective To explore the accuracy of electronic nose(E-nose)breath analysis for the early detection of colorectal cancer(CRC)based on various machine learning algorithms,and its preliminary ability to identify common traditional Chinese medicine(TCM)syndromes associated with CRC.Methods The study was designed as a prospective,observational research.A total of 125 CRC patients and 107 healthy subjects were recruited.Four diagnostic information were collected,and TCM syndrome distribution was analyzed using syndrome differentiation methods.Exhaled breath profiles were collected using the Cyranose320 portable electronic nose,and five machine learning algorithms-random forest,k-nearest neighbors,logistic regression,support vector machine,and eXtreme gradient boosting-were employed to identify the exhaled breath profiles related to CRC and its common disease locations and pathogenic characteristics.Results①The common disease location syndromes for CRC,from high to low prevalence,were the large intestine,spleen,liver,and stomach,while the disease nature syndromes,from high to low,were Qi deficiency,Yang deficiency,Qi stagnation,and Yin deficiency.②The random forest model effectively distinguished CRC patients from healthy subjects,achieving an accuracy of 88.57%,specificity of 80.56%,sensitivity of 97.06%,and an AUC of 0.89(95%CI:0.81-0.98).③The random forest model exhibited the best performance in identifying the exhaled breath profiles of common CRC disease location and nature syndromes,with classification accuracy generally exceeding 90.00%.Conclusion The E-nose exhaled breath profiles not only enable early non-invasive detection of CRC but also provide objective,data-driven evidence for the diagnosis of TCM syndromes associated with CRC.
作者
谭施言
由凤鸣
王倩
任益锋
王巧灵
王钧冬
TAN Shiyan;YOU Fengming;WANG Qian;REN Yifeng;WANG Qiaoling;WANG Jundong(Hospital of Chengdu University of Traditional Chinese Medicine,Chengdu 610075,China)
出处
《世界科学技术-中医药现代化》
北大核心
2026年第1期97-106,共10页
Modernization of Traditional Chinese Medicine and Materia Medica-World Science and Technology
基金
四川省科学技术厅自然科学基金面上项目(2022NSFSC0670):基于E-nose呼气图谱的结肠癌“证”转归模型研究,负责人:王巧灵。
关键词
结直肠癌
电子鼻
呼气图谱
机器学习
中医证素
Colorectal cancer
Electronic nose
Exhaled breath profiles
Machine learning
Traditional Chinese medicine(TCM)syndrome elements
