床旁超声影像组学在诊断新生儿吸入性肺炎轻症与重症中的价值研究  

The Value of Bedside Ultrasound-based Radiomics Model in the Diagnosis of Mild and Severe Neonatal Aspiration Pneumonia

作　　者：牛慧敏[1] 刘欣[1] 于明月[1] Niu Huimin;Liu Xin;Yu Mingyue(Department of Ultrasound,Hebei General Hospital,Shijiazhuang 050051,China)

机构地区：[1]河北省人民医院超声科,石家庄市050051

出　　处：《中国超声医学杂志》2023年第8期871-874,共4页Chinese Journal of Ultrasound in Medicine

基　　金：河北省卫计委重点课题计划(No.20170344)。

摘　　要：目的基于床旁超声影像组学特征和临床一般指标建立联合预测模型,探究其在诊断新生儿吸入性肺炎轻症与重症中的价值。方法回顾性分析根据超声影像学与实验室指标确诊的181例吸入性肺炎患儿资料,其中轻度吸入性肺炎75例,重度吸入性肺炎106例。使用itk-snap软件进行感兴趣区域(ROI)勾画,基于Python的Pyradiomics包进行特征提取。对181例超声影像资料进行6∶4随机切割,分为训练集(n=109)、验证集(n=72)。分别对训练集和验证集进行标准化(Z-score)及单因素、多因素分析,建立影像组学模型,然后分别在训练集和验证集中评估和验证模型。采用受试者工作特征(ROC)和校准曲线评估模型的诊断价值。结果训练集和验证集中超声影像组学特征曲线下面积(AUC)分别为0.932、0.947。校准曲线提示超声影像组学模型在训练集和验证集中均具有较好的校准能力。结论床旁超声影像组学在诊断新生儿吸入性肺炎严重程度中具有潜在价值。Objective A combined prediction model was established based on bedside ultrasound imaging characteristics and general clinical indicators to explore its value in the diagnosis of mild and severe aspiration pneumonia in neonates.Methods The data of 181children with aspiration pneumonia confirmed by ultrasound imaging and laboratory indicators were analyzed retrospectively,including 75cases with mild aspiration pneumonia and 106cases with severe aspiration pneumonia.itk-snap software was used for ROI delineation,and Pyradiomics package based on Python was used for feature extraction.A total of 181cases of ultrasound image data were randomly cut 6∶4into a training set(n=109)and a validation set(n=72).Standardization(Z-score)and univariate and multivariate analysis were conducted on the training set and validation set respectively to establish the radiomics signature,and then the model was evaluated and verified in the training set and validation set respectively.The diagnostic value of the model was evaluated by ROC and calibration curve.Results The AUC values of ultrasound imaging in training set and validation set were 0.932and 0.947respectively.The calibration curve indicates that the ultrasound image-based machine learning model had good calibration ability in both the training and validation sets.Conclusions Bedside ultrasound-based radiomics model has potential value in diagnosing the severity of neonatal aspiration pneumonia.

关 键 词：床旁超声 影像组学 新生儿 吸入性肺炎 严重程度 

分 类 号：R445.1[医药卫生—影像医学与核医学] R722.135[医药卫生—诊断学]