Document resource
Introduction Systemic Sclerosis (SSc) often leads to a progressive and highly fatal form of interstitial lung disease (ILD). While high-resolution CT (HRCT) is the gold standard for diagnosis, its manual scoring is prone to variability. Recent advancements in AI offer a promising solution for standardized and reproducible imaging analysis. This study aimed to correlate an AI-based HRCT score with lung function, assess treatment impact on disease progression, and identify early predictors of treatment efficacy.Material and Methods This was a retrospective study of SSc patients with ILD treated at our rheumatology department between 2006 and 2024. Patients were included if they had at least two HRCT scans, serial pulmonary function tests, and a minimum three-year follow-up. HRCT scans were analyzed using Contextflow ADVANCE Chest CT, an AI software that quantifies radiologic abnormalities and generates a weighted lung abnormality score (QILD). Statistical analyses were performed using Prism GraphPad software.Results From 450 ssc patients, 67 met the inclusion criteria ( table 1). Of these, 26 (39%) died during the study period. At baseline, 85% (57) of patients showed significant lung anomalies, primarily ground glass opacities, with a median QILD score of 22 (range 8-69). Progressive ILD was observed in 39% (26) of patients, while 33% (22) showed an improvement in their QILD score.A significant correlation was found between baseline QILD scores and lung function tests (FVC and TLCO) performed at least three years later (p=0.029 for both). However, we observed no correlation between the type of treatment and the change in QILD score on HRCT scans 1-2 years after treatment initiation.Furthermore, a baseline QILD score of >20 was associated with a significantly higher 5-year mortality rate (19.5%, 8/41) compared to a score of <20 (4.8%, 2/42), a difference that was statistically significant (p=0.0481).Conclusions Our findings suggest that a higher baseline QILD score in SSc patients is a valuable prognostic marker for both future lung function decline and increased 5-year mortality risk. While a correlation between baseline scores and long-term outcomes was evident, the lack of a short-term link between specific treatments and QILD score changes warrants further investigation. This highlights the complexity of predicting ILD progression in SSc and the potential utility of AI-based tools in this context.Abstract P.119 Table 1Demographic and clinical data