Background: Interstitial lung disease (ILD) represents a major complication of systemic sclerosis (SSc) and is a leading cause of morbidity and mortality in affected patients. Increasing interest has been shown in airway abnormalities associated with ILD. In patients with SSc, bronchiectasis is reported in up to two-thirds of those with ILD, primarily as a consequence of fibrotic distortion of lung architecture. Additional contributing factors may include aspiration due to esophageal dysmotility, as well as the modifying effects of infections or immunosuppressive therapies. The presence of bronchiectasis has been reported to be prognostically unfavorable in these patients. A proposed pathogenic mechanism involves excessive accumulation of collagen and extracellular matrix in the bronchial smooth muscle and elastin layers, leading to airway fibrosis and narrowing. Furthermore, the recently updated 2025 ERS/ATS classification of interstitial pneumonias has introduced a new morphological ILD pattern, termed bronchiolocentric interstitial pneumonia (BIP). This pattern may occur in isolation or in combination with other ILD patterns, such as nonspecific interstitial pneumonia (NSIP) or usual interstitial pneumonia (UIP), and is characterized by airway-centered fibrosis and inflammation extending into the adjacent peribronchiolar interstitium. The clinical and prognostic significances of BIP in SSc-associated ILD remain largely unknown.

Airway morphology is evaluated, at best, using semiquantitative scoring systems based on HRCT. More recently, artificial intelligence (AI)–based analysis of HRCT has emerged as a novel tool for the objective, reproducible, and reliable assessment of pulmonary diseases.

Objectives: The aim of this study is to correlate AVIEW measures of airways, a Deep Learning based software for HRCT image assessment, with ILD-progression and disease-related mortality in SSc patients.

Methods: The AVIEW software (Coreline Soft, South Korea; CE-marked) was used to analyze airway structures on HRCT images from a cohort of consecutive patients with SSc-ILD, evaluated at baseline and after 24±3 months. AVIEW enables automated airway segmentation and extraction of quantitative airway measurements. Airway metrics included an overall bronchiectasis score and a composite quantitative bronchiectasis score reflecting both disease extent and severity. Additional parameters comprised airway wall volume, lumen volume, and standardized measures of wall thickness, including wall area percentage, and WT-Pi10. These measurements were assessed globally and at the 5th, 6th, 7th, and 8th bronchial generation levels. Baseline airway metrics were evaluated for their association with ILD progression over a 24-month period. ILD progression was defined using clinical, functional, and imaging criteria derived from the INBUILD study. Furthermore, longitudinal changes in AVIEW-derived airway measurements between consecutive HRCT examinations over 24 months were analyzed for their association with SSc-related mortality during the subsequent 36 months. To reduce the risk of overfitting in predictive analyses, a bootstrap resampling procedure with 2,000 iterations was applied.

Results: A total of 140 HRCT scans from 70 patients with SSc-ILD were analyzed (mean age 57.7 ± 14.2 years; males 17.1%; diffuse cutaneous subset 52.9%). ILD progression occurred in 33 patients (47.1%). Compared with non-progressors, these patients showed significantly higher airway wall volume at the 5th-8th bronchial generations (p < 0.05), without differences in lumen-related measures or bronchiectasis scores. Among standardized metrics, progressors also exhibited higher wall area percentage at the 5th -6th bronchial generations (all p < 0.05). A prediction model including wall volume at the 5th bronchial generation together with conventional assessment tools (forced vital capacity [FVC], diffusing capacity for carbon monoxide [DLCO], and ILD extent based on visual HRCT assessment) demonstrated the ability to predict INBUILD-defined ILD progression (AUC = 0.693). This AI-integrated model showed a superior performance compared with a model based on conventional assessment tools alone (likelihood ratio test p = 0.001). During the 36 months following the follow-up HRCT, 11 SSc-related deaths were observed. In multivariable models adjusted for baseline age and the occurrence of INBUILD progression between the two HRCT examinations, increases in bronchiectasis score, wall volume at the 7th-8th bronchial generations, lumen volume at the 6th–8th generations, and WT-Pi10 at the 5th–7th generations were independently associated with SSc-related mortality (all p < 0.05).

Conclusions: AI-based airway evaluation provides a useful tool for prognostic stratification in SSc-ILD. These findings highlight, on the one hand, the added value of this technology in disease assessment and, on the other, the significant role of airway involvement in the pathogenesis of SSc-ILD. Moreover, luminal and wall-related distal airway changes convey distinct and complementary information, underscoring the different pathological significance of bronchiectasis-related luminal dilatation and airway wall thickening related to fibrotic involvement of bronchial and peribronchial tissues. This study was funded by the European Union - NextGenerationEU (Piano Nazionale di Ripresa e Resilienza PNRR-MCNT2-2023-12377894). Project title: “ Early recognition of progressive lung fibrosis in systemic rheumatic diseases:a characterization of the pulmonary environment through extracellular vesicles, advanced and functional imaging”, CUP: C53C23001060007.

REFERENCES: NIL.

Acknowledgments: NIL.

Disclosure of Interests: None declared.