The innovative diagnostic model facilitates the differentiation between non - tuberculous mycobacterial lung disease and pulmonary tuberculosis

TL;DR

This study developed a diagnostic model to distinguish between two lung diseases using CT scans, symptoms, and clinical data, achieving high accuracy.

Contribution

A novel diagnostic model combining clinical and imaging data for differentiating NTM-LD and PTB-LD is proposed.

Findings

Key discriminative factors included cough with sputum, hemoptysis, and imaging features like thin-walled cavities.

The model achieved 82.5% sensitivity and 85.5% specificity in training, and 78% sensitivity and 85% specificity in validation.

Abstract

To construct a differential diagnostic model for Non-Tuberculous Mycobacterial Lung Disease (NTM-LD) and Pulmonary Tuberculosis Lung Disease (PTB-LD).

Retrospective analysis of 300 NTM-LD and 300 PTB-LD patients (pathogen-confirmed) was performed. Patients were randomly split into training (2/3) and validation (1/3) sets. CT imaging, clinical data, and symptoms were analyzed. Logistic regression identified significant discriminative features, followed by random forest modeling to develop a diagnostic tool with web-based calculator. Model performance was validated using the independent validation set.

Univariate and multivariate analyses identified key discriminative factors (P<0.05): cough with sputum, hemoptysis, thin-walled cavities, centrilobular nodules, bronchiectasis, diabetes, and autoimmune diseases. The diagnostic model achieved 82.5% sensitivity and 85.5% specificity (ROC analysis), with validation showing 78% sensitivity and 85% specificity, confirming strong discriminative power and calibration.

The model constructed based on patients’ CT imaging, basic clinical data, and symptomatic signs demonstrates commendable performance in the differential diagnosis of NTM-LD and PTB-LD, offering a convenient and practical auxiliary tool for clinical practice.

Figures