【摘要】 目的根据肝癌临床诊断的需求,建立肝癌诊断预测模型,以达到无创检测肝癌的目的。方法利用德国企业产ILD3000型电子鼻设备采集正常受试者和肝癌患者的呼气数据,对呼气所得时间序列数据进行特征提取,包括序列数据的最大值、最小值、均值、标准差、序列数据总和等统计学特征。结合特征降维算法和机器学习分类模型对呼气特征数据进行正常受试者和原发性肝癌患者的二分类实验。结果通过模型选择和参数调整,在线性核函数支持向量机上对呼气数据取得923%的最优二分类结果。结论以正常受试者和肝癌患者的呼气数据为样本,利用机器学习建模的方法可以对肝癌做出诊断预测,且在此数据上,线性核函数支持向量机算法具有最好的分类效果。更多还原

【Abstract】 Objective To establish a predictive model for diagnosis of liver cancer in order to achieve non-invasive detection of liver cancer.MethodsThe exhalation data of normal subjects and patients with liver cancer were collected by ILD.3000 electronic nose equipment manufactured by German enterprises. The features of exhalation time series data were extracted,including maximum,minimum,mean,standard deviation and sum of sequence data. Combining feature dimension reduction algorithm and machine learning classification model,a binary classification experiment was carried out for normal subjects and patients with primary liver cancer using expiratory feature data.ResultsThrough model selection and parameter adjustment,92.3% of the optimal binary classification results were obtained for exhalation data on linear kernel function support vector machine.ConclusionsUsing the exhalation data of normal subjects and patients with liver cancer as samples, the diagnosis and prediction of liver cancer can be made by using machine learning modeling method. On this data,the linear kernel function support vector machine algorithm has the best classification effect.更多还原