Back to all papers

Precise discrimination of mycobacterial pulmonary diseases via multimodal machine learning integrating chest CT and clinical markers.

July 9, 2026pubmed logopapers

Authors

Jin Y,Ding J,Ouyang J,Yao Z,Wang L,Xu R,Jin X

Affiliations (5)

  • Department of General Internal Medicine, Zhejiang Cancer Hospital, Hangzhou, Zhejiang, China.
  • Department of Respiratory and Critical Care Medicine, The People's Hospital of Yuhuan, Yuhuan, China.
  • Department of Respiratory and Critical Care Medicine, The First Affiliated Hospital of Wenzhou Medical University, Wenzhou, China.
  • Wenzhou Institute, University of Chinese Academy of Sciences, Wenzhou, China.
  • Shanghai Key Lab of Modern Optical System, Terahertz Spectrum and Imaging Technology Cooperative Innovation Center, Terahertz Technology Innovation Research Institute, University of Shanghai for Science and Technology, Shanghai, China.

Abstract

Differentiating <i>Mycobacterium tuberculosis lung disease</i> (MTB-LD) from nontuberculous mycobacterial lung disease (NTM-LD) remains clinically challenging because of overlapping symptoms and imaging features, as well as the slow turnaround of conventional culture-based diagnostics. This retrospective study enrolled 102 patients with microbiologically confirmed mycobacterial lung disease, including 53 patients with MTB-LD and 49 patients with NTM-LD. We developed and validated an interpretable multimodal machine-learning framework integrating clinical symptoms, hematological biomarkers, and high-resolution computed tomography (HRCT) features. Three representative classifiers, including k-nearest neighbors, logistic regression, and random forest, were used to evaluate the discriminative contribution of different feature modalities. Multimodal integration of HRCT, clinical, and laboratory features showed better discriminative performance than single-modality approaches. Among the three representative classifiers, the random forest model achieved the best hold-out test performance, with an AUC of 0.92, sensitivity of 0.89, specificity of 0.93, and F1-score of 0.90. Key predictive contributors included cystic bronchiectasis, tree-in-bud sign, fever, and selected laboratory biomarkers. These findings suggest that routinely available multimodal clinical data may provide preliminary decision support for MTB-LD/NTM-LD differentiation. However, the proposed framework should be regarded as an exploratory decision-support tool, and external validation is required before clinical implementation.

Topics

Journal Article

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.