基于机器学习的支原体肺炎诊断中临床表现与生化指标的关键特征分析

Key feature analysis of clinical manifestations and biochemical indicators in mycoplasma pneumonia diagnosis based on machine learning

支原体肺炎病例的诊断过程中涉及大量的临床指标和生物化学指标,由于没有统一的单一确诊指标,医生常常面临如何正确诊断的困扰。为获得支原体肺炎诊断和评估中的关键临床表现指标和生物化学指标,采用多种机器学习分类模型进行定量分析。首先,通过机器学习方法对21项临床指标和22项生物化学指标进行训练,通过计算模型的准确率(Accuracy)、精度(Precision)、召回率(Recall)和F1分数,获得可以准确进行是否患有支原体肺炎的分类机器学习模型;在保证模型可靠性的基础上,进行特征重要性评估,然后对重要特征进行筛选和分析。综合结果分析表明,在临床表现数据中,Three Concave Sign.1、Cough Nature、Moist Crackles、Rhonchus or Wheeze以及Pleural Effusion指标对模型分类能力的影响最为显著,在生物化学数据中,Nucleic Acid PCR、IgM(acute stage)、White blood cell(WBC)、CRP(C Reaction Protein)以及Lymphocyte%(L)这五项指标对模型分类的影响最大。该研究为支原体肺炎的关键特征筛选和分析提供了关键指标,为临床诊断和评估提供了基础参考。

The diagnostic process of mycoplasma pneumoniae pneumonia(MPP)cases involves a large number of clinical and biochemical indicators.Clinical physicians often face with challenges in correctly diagnosing MPP due to the lack of well‑defined diagnostic indicators.In order to obtain the key clinical indicators and biochemical indicators in the diagnosis and assessment of MPP,this study employed machine learning classification models for quantitative analysis.Firstly,machine learning models were trained using 21 clinical indicators and 22 biochemical indicators.The models’accuracy,precision,recall,and F1 scores were computed to examine the performance of machine learning models for classifying MPP.Machine learning models have achieved the accuracy of 0.94 and 0.78 in clinical manifestations and biochemistry indicators,respectively.After ensuring the reliability of the models,feature importance assessment is conducted to select important features.The comprehensive result analysis showed that among the clinical manifestations data,Three Concave Sign.1,Cough Nature,Moist Crackles,Rhonchus or Wheeze,and Pleural Ef‑fusion indicators contributed the most to the model classification ability.In the biochemical data,Nucleic Acid PCR,IgM(acute stage),White blood cell(WBC),CRP(C Reaction Protein),and Lymphocyte%(L)showed the greatest influence on the model's classification.This study provides crucial indicators for the diagnosis of MPP,serving as a foundational reference for clinical diag‑nosis and assessment.