影像科学与光化学2024，Vol.42Issue(4)：371-377,7.DOI:10.7517/issn.1674-0475.2024.04.11
基于定量CT对难治型支原体肺炎ANN模型的构建
Construction of ANN Model of Refractory Mycoplasma Pneumonia Based on Quantitative CT
张琳清 1呼甜 2何思怡 1李佳颖 1路紫旸 1王一涵 1王雷 2黄晓旗2
作者信息
- 1. 延安大学,陕西延安 716000
- 2. 延安大学附属医院影像科,陕西延安 716000
- 折叠
摘要
Abstract
Objective:Through the establishment of artificial neural network(ANN)model based on clinical index and quantitative imaging index,to explore the value of clinical features combined with quantitative CT index in the prediction of early refractory Mycoplasma pneumoniae pneumonia(RMPP).Methods:The clinical and imaging data of 116 children with Mycoplasma pneumoniae pneumonia(MPP)diagnosed in Yan'an University Affiliated Hospital from September 2023 to November 2023 were collected retrospectively,including 33 cases of RMPP and 83 cases of general Mycoplasma pneumoniae pneumonia(GMPP).The laboratory examination and CT scan images of the children were collected.Computer-aided software was used to conduct quantitative analysis on the children's chest CT to obtain quantitative indicators.The laboratory indexes and quantitative CT indexes were analyzed by 2 test or Mann-Whitney U test,the indexes with significant differences were incorporated into the artificial neural network model,the prediction efficiency of the model was evaluated by ROC curve,and the prediction importance of all indexes was ranked.Results:Analysis of the difference in clinical symptoms between the two groups showed that the RMPP group(4.5 years old)was older than the GMPP group(4 years old)and had a longer hospitalization time(RMPP=11.5 days)(P＜0.05).Difference analysis of laboratory indicators showed that the CRP(14.63 mg/L),LDH(312 U/L),and neutrophil ratio(65.2％)in the peripheral serum of children in the RMPP group were higher than those in the GMPP group,while the lymphocyte ratio(26.3％)was lower than that in the GMPP group.The IgE(26.35 IU/mL)of children with RMPP was lower than that of GMPP(P＜0.01).The quantitative parameters of LeV(32.92 mL),LeV％(6.09％),MLeD(-242.77 HU),LM(33.13 g)in the RMPP group's CT scans were higher than those in GMPP group.GGO(-300)(0.43％)was lower than GMPP.After the two are included in the model,the predicted AUC of the model is 0.968.The importance of LM,LeV％,LeV,neutrophil count and ratio,and LDH normalization are all more than 50％.Conclusion:The ANN model constructed by quantitative CT index combined with clinical characteristics has a good ability to predict RMPP in children.关键词
支原体肺炎/预测模型/计算机体层成像Key words
Mycoplasma pneumoniae pneumonia/prediction model/computed tomograph(CT)分类
化学化工引用本文复制引用
张琳清,呼甜,何思怡,李佳颖,路紫旸,王一涵,王雷,黄晓旗..基于定量CT对难治型支原体肺炎ANN模型的构建[J].影像科学与光化学,2024,42(4):371-377,7.
