AI-based CT quantification reveals small airway loss and vascular simplification in rheumatoid arthritis-associated lung disease.
Authors
Affiliations (5)
Affiliations (5)
- School of Medical Technology, Shaanxi University of Chinese Medicine, Xian Yang, 712046, China.
- School of Medical Technology, Shaanxi University of Chinese Medicine, Xian Yang, 712046, China. [email protected].
- Department of Medical Imaging, Affiliated Hospital of Shaanxi University of Chinese Medicine, 2 Weiyang Western Road, Xian Yang, 712000, China. [email protected].
- Department of Rheumatology Immunohematology, Affiliated Hospital of Shaanxi University of Chinese Medicine, Xian Yang, China.
- Department of Medical Imaging, Affiliated Hospital of Shaanxi University of Chinese Medicine, 2 Weiyang Western Road, Xian Yang, 712000, China.
Abstract
To investigate pulmonary structural changes in patients with rheumatoid arthritis (RA) using an artificial intelligence (AI)-based CT segmentation model and evaluate the diagnostic potential of quantitative imaging biomarkers. This cross-sectional study included 556 non-smoking RA patients and 472 non-smoking healthy controls. An AI-based CT segmentation model was used to quantify total lung volume (TLV), percentage of low attenuation area (LAA%-200-700), total airway volume (TAV), small airway volume (TAV2mm), total vessel volume (TVV), small vessel volume (TVV5mm), vessel tortuosity (VT), and fractal dimension (FD). Diagnostic performance was assessed using receiver operating characteristic (ROC) curve analysis. Subgroup analysis compared RA patients with and without respiratory symptoms. RA patients exhibited significantly lower TLV, increased LAA%-200-700, marked reductions in small airway volume (TAV2mm) and small vessel volume (TVV5mm), and a minimal, albeit statistically significant, alteration in VT compared to healthy controls (all p < 0.001). TAV2mm demonstrated the highest single-parameter diagnostic performance (AUC: 0.801). The combination of TAV2mm, TVV5mm, and LAA%-200-700 achieved superior discrimination (AUC: 0.837). Symptomatic RA patients showed greater reductions in vascular volumes and more extensive interstitial abnormalities than asymptomatic patients (p < 0.05), confirming clinical relevance. RA patients exhibit quantifiable small airway loss and imaging features suggestive of vascular simplification. A combination of small airway, small vessel, and interstitial metrics effectively distinguishes RA patients from healthy controls. Key Points • Potential Imaging Signature: AI-based CT quantification identifies a potential "vascular simplification" pattern (primarily characterized by reduced TVV5mm) in RA lungs, which appears distinct from the vascular changes seen in classic pulmonary hypertension and warrants further validation. • Dual-Compartment Damage: RA patients exhibit concurrent loss of small airways (TAV2mm, AUC=0.801) and pruning of small vessels (TVV5mm), with both abnormalities correlating with interstitial changes (LAA%-200-700). • Multi-parametric Assessment: The combination of small airway, small vessel, and interstitial metrics achieves excellent discrimination (AUC=0.837), outperforming any single parameter for detecting RA associated lung involvement.