基于3种细胞因子的尘肺病筛查支持向量机模型的建立

Establishment of support vector machine model for pneumoconiosis screening based on three cytokines

目的:建立基于转化生长因子β1(TGF-β1.、血小板源性生长因子(PDGF)、结缔组织生长因子(CTGF)的支持向量机模型(SVM)用于尘肺病的筛查。方法:选择70例男性尘肺病患者(尘肺病组),77例体检健康的男性(对照组),分别采集外周血并分离血清。采用ELISA法检测血清中TGF-β1、CTGF、PDGF的含量。采用SPSS Clementine软件分别构建Fisher判别分析模型和SVM模型,比较2种模型诊断尘肺病的效能。结果:基于血清TGF-β1、PDGF、CTGF含量建立的Fisher判别分析模型诊断尘肺病的准确度、灵敏度、特异度分别为78.1%、95.0%、61.9%,而SVM模型的准确度、灵敏度、特异度分别为87.8%、95.0%、81.0%;SVM模型的AUC为0.908,优于Fisher判别分析模型(0.830)(Z=3.181,P=0.002.。结论:建立了基于人血清TGF-β1、PDGF、CTGF含量、可用于尘肺病筛查的SVM模型,且筛查效果较好。

Aim:To establish a support vector machine(SVM)model for screening pneumoconiosis based on transforming growth factor-β1(TGF-β1),platelet derived growth factor(PDGF),and connective tissue growth factor(CTGF).Methods:Seventy males with pneumoconiosis(pneumoconiosis group)and 77 healthy males(control group)were selected.Peripheral blood of the subjects was collected,and the serum was separated.The contents of TGF-β1,CTGF,and PDGF were determined by ELISA.SPSS Clementine software was used to construct Fisher discrimination analysis model and SVM model.The efficacy of 2 models for screening pneumoconiosis was compared.Results:The accuracy,sensitivity,and specificity of Fisher discrimination analysis model in diagnosis of pneumoconiosis were 78.1%,95.0%,61.9%,and those for SVM model were 87.8%,95.0%,81.0%,respectively.The AUC of SVM model was 0.908,higher than that(0.830)of Fisher discrimination analysis model(Z=3.181,P=0.002..Conclusion:SVM model based on serum TGF-β1,CTGF,and PDGF has been established with better prediction efficacy in pneumoconiosis screening.