文献速递胸部CT

探索肺纹理特征在区分寻常型与非特异性间质性肺炎中的价值

Exploring the Value of Features of Lung Texture in Distinguishing Between Usual and Nonspecific Interstitial Pneumonia.
PubMed 文献 · Acad Radiol · 2022-07-14

回顾性分析96例间质性肺炎(寻常型40例、非特异性56例),利用肺智能工具包分割肺并提取纹理特征,经方差分析、LASSO和多因素logistic回归建模区分两型。训练组与测试组AUC分别为0.952和0.838,敏感度0.821和0.667,特异度0.949和0.824。

要点

  • 从全肺提取100个纹理特征,降维后保留8个特征用于建模。
  • 训练组AUC 0.952、敏感度0.821、特异度0.949。
  • 测试组AUC 0.838、敏感度0.667、特异度0.824。

核心结论

  • 利用CT图像获得的肺纹理特征有可能区分寻常型间质性肺炎与非特异性间质性肺炎。

研究设计与样本

  • 回顾性分析96例间质性肺炎(寻常型40例、非特异性56例),所有患者均行CT扫描,采用肺智能工具包进行肺分割和纹理特征提取,经方差分析、LASSO和多因素logistic回归筛选变量并建立模型,以ROC曲线、AUC、敏感度和特异度评价模型性能。

诊断标准与关键数值

  • 多因素logistic回归模型训练组AUC 0.952、敏感度0.821、特异度0.949。
  • 多因素logistic回归模型测试组AUC 0.838、敏感度0.667、特异度0.824。
  • 从全肺提取100个纹理特征,特征降维后保留8个特征。

局限性

  • 原文未报告局限性。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

英文原文(摘要)· 点击展开

RATIONALE AND OBJECTIVES: This article aims to explore the potential use of lung texture assessed in CT images in distinguishing between the usual interstitial pneumonia and the nonspecific interstitial pneumonia.

MATERIALS AND METHODS: A retrospective analysis of 96 cases of interstitial pneumonia was performed. Among these cases, there were 40 cases of usual interstitial pneumonia (UIP) and 56 cases of the nonspecific interstitial pneumonia (NSIP) . All of the patients underwent computed tomography (CT) scans. A lung intelligence kit (LK) was utilized to perform lung segmentation and texture feature extraction. The significant variables were determined by variance analysis, least absolute shrinkage and selection operator (LASSO) and multivariate logistic regression. Finally, a multivariate logistic regression model was established to distinguish between the two types of interstitial pneumonia. Receiver operating characteristic (ROC) curves, area under the curve (AUC) values, sensitivity, and specificity were used to evaluate the performance of the established model.

RESULTS: A total of 100 texture features were extracted from the whole lung that was segmented by LK, and 8 features remained after feature reduction. The AUC, sensitivity, and specificity of the multivariate logistic regression model in the training group and the test group were 0.952 and 0.838, 0.821 and 0.667, and 0.949 and 0.824, respectively.

CONCLUSION: It is possible to distinguish between UIP and NSIP using lung texture features obtained from CT images.

来源:Acad Radiol,查看原文。版权归原作者所有。

引用

  1. Chen, Cheng, Yang, et al. Exploring the Value of Features of Lung Texture in Distinguishing Between Usual and Nonspecific Interstitial Pneumonia. Acad Radiol. 2022. PMID: 35843833. doi:10.1016/j.acra.2022.06.011

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。