BMC Medical Imaging (Oct 2022)

A CT-based nomogram for differentiating invasive fungal disease of the lung from bacterial pneumonia

  • Meilin Gong,
  • Jingmei Xu,
  • Kang Li,
  • Ke Li,
  • Yuwei Xia,
  • Yang Jing,
  • Jiafei Chen,
  • Jing Li,
  • Jing Yang,
  • Mingshan Du,
  • Wenjing Hou,
  • Yuan Ou,
  • Lian Li,
  • Wei Chen

DOI
https://doi.org/10.1186/s12880-022-00903-5
Journal volume & issue
Vol. 22, no. 1
pp. 1 – 11

Abstract

Read online

Abstract Background There is an annual increase in the incidence of invasive fungal disease (IFD) of the lung worldwide, but it is always a challenge for physicians to make an early diagnosis of IFD of the lung. Computed tomography (CT) may play a certain role in the diagnosis of IFD of the lung, however, there are no specific imaging signs for differentiating IFD of lung from bacterial pneumonia (BP). Methods A total of 214 patients with IFD of the lung or clinically confirmed BP were retrospectively enrolled from two institutions (171 patients from one institution in the training set and 43 patients from another institution in the test set). The features of thoracic CT images of the 214 patients were analyzed on the picture archiving and communication system by two radiologists, and these CT images were imported into RadCloud to perform radiomics analysis. A clinical model from radiologic analysis, a radiomics model from radiomics analysis and a combined model from integrating radiologic and radiomics analysis were constructed in the training set, and a nomogram based on the combined model was further developed. The area under the ROC curve (AUC) of the receiver operating characteristic (ROC) curve was calculated to assess the diagnostic performance of the three models. Decision curve analysis (DCA) was conducted to evaluate the clinical utility of the three models by estimating the net benefit at a range of threshold probabilities. Results The AUCs of the clinical model for differentiating IFD of lung from BP in the training set and test sets were 0.820 and 0.827. The AUCs of the radiomics model in the training set and test sets were 0.895 and 0.857. The AUCs of the combined model in the training set and test setswere 0.944 and 0.911. The combined model for differentiating IFD of lung from BP obtained the greatest net benefit among the three models by DCA. Conclusion Our proposed nomogram, based on a combined model integrating radiologic and radiomics analysis, has a powerful predictive capability for differentiating IFD from BP. A good clinical outcome could be obtained using our nomogram.

Keywords