基于一致性预测算法的病毒性脑膜炎辅助诊断模型研究  

Research on auxiliary diagnosis model of viral meningitis based on conformal prediction algorithm

作　　者：吴能光 陈拓 陈虹[2] WU Nengguang;CHEN Tuo;CHEN Hong(Information Department,the First Affiliated Hospital of Fujian Medical University,Fuzhou 350005,Fujian Province,China;Pathology Department,the First Affiliated Hospital of Fujian Medical University,Fuzhou 350005,Fujian Province,China;Information Center of Zhejiang Provincial People's Hospital)

机构地区：[1]福建医科大学附属第一医院信息中心,福州350005 [2]福建医科大学附属第一医院病理科,福州350005 [3]浙江省人民医院信息中心

出　　处：《中国数字医学》2025年第4期84-88,共5页China Digital Medicine

摘　　要：目的:构建病毒性脑膜炎置信分类模型,为临床决策提供参考价值。方法:以病毒性脑膜炎数据为研究样本,采用神经网络(NN)和K最近邻(KNN)作为一致性预测器(CP)的奇异映射底层算法构建病毒性脑膜炎置信模型,即CP-NN和CP-KNN。结果:在设定置信度水平为0.998,CP-NN模型表现优异,其准确率为0.964,精确率为0.969,召回率为0.981,调和平均数为0.975。结论:CP-NN作为病毒性脑膜炎疾病辅助诊断模型,能够解决诊断中个性化高风险评估问题和传统机器学习模型退化问题,同时预测结果附带置信度,使得病毒性脑膜炎诊断更符合医疗需求。Objective To establish a confidence classification model of virus meningitis,so as to provide valuable reference for clinical decision-making.Methods The viral meningitis data were utilized as research samples in this study,with the singular mapping underlying algorithm of Neural Network(NN)and K-Nearest Neighbor(KNN)serving as Conformal Prediction(CP),so as to establish the confidence model of viral meningitis,namely CP-NN and CP-KNN.Results The CP-NN model performed excellently when the confidence level was set to 0.998,achieving the accuracy of 0.964,the precision of 0.969,the recall rate of 0.981,and the F1 score of 0.975.Conclusion As an auxiliary diagnostic model for viral meningitis disease,CP-NN can address the issue of personalized high-risk assessment and the degradation of traditional machine learning model in diagnosis.Additionally,the prediction results are accompanied with confidence,aligning the diagnosis of viral meningitis more closely with medical needs.

关 键 词：一致性预测器 病毒性脑膜炎 临床辅助诊断 置信分类模型 

分 类 号：R197.3[医药卫生—卫生事业管理] R319[医药卫生—公共卫生与预防医学]