CO:07:2 | Quantitative computed tomography analysis of pulmonary disease in systemic sclerosis: correlation between radiomics and pulmonary function tests

Francesco Pollastri¹, Loredana Carobene², Giovanni Adami¹, Michele Pernigotto³, Federico Aldegheri¹, Angela Ventura³, Claudio Micheletto², Susanna Baltieri³, Marco Sebastiani^4|6, Andreina Manfredi⁵, Davide Gatti¹, Maurizio Rossini¹, Angelo Fassio¹. | ¹UOC Reumatologia, Università degli Studi di Verona; ²UOC Pneumologia, Università degli Studi di Verona; ³ UOC Radiologia, Università degli Studi di Verona; ⁴Universita degli Studi di Parma; ⁵UUOC Reumatologia, Università degli Studi di Modena e Reggio Emilia; ⁶UOC Reumatologia, Azienda Unita' Sanitaria Locale, Piacenza, Italy

Background and objectives. High-resolution computed tomography (HRCT)-based radiomic analysis is gaining attention as a tool for quantifying interstitial lung disease (ILD) in systemic sclerosis (SSc). Pulmonary function tests (PFTs) are standard for functional assessment but have limited sensitivity in early disease stages. We aimed to explore the potential of HRCT-derived radiomic features in SSc patients with and without ILD.

Methods. In this cross-sectional study, we enrolled SSc subjects who underwent both HRCT and PFTs. Radiomic parameters—skewness, kurtosis, high-attenuation areas (pHAA-700 and pHAA-600;-250), and lung volume—were extracted. Participants were classified into four groups based on HRCT pattern: no ILD, non-fibrosing NSIP (nfNSIP), fibrosing NSIP (fNSIP), and usual interstitial pneumonia (UIP). ANOVA with post-hoc comparisons, linear regressions, and ROC analyses were conducted to assess diagnostic performance and associations with PFTs.

Results. Ninety-three subjects were enrolled. All radiomic parameters significantly differed across ILD patterns (p<0.001 for skewness, kurtosis, pHAA metrics; p=0.005 for lung volume). PFT values also varied among subgroups (ppFVC p=0.005; ppTLC p<0.001; ppDLCO p=0.004). Radiomic parameters were moderately correlated with ppFVC, ppTLC, and ppDLCO in ILD patients (R ranging in absolute terms from 0.33 to 0.58, p<0.001), with significant interaction terms (PFT parameter*presence of ILD). ROC analysis showed that skewness (AUC 0.897, p=0.0073) and kurtosis (AUC 0.890, p=0.010) outperformed ppTLC (AUC 0.744) in identifying ILD.

Conclusions. Radiomic analysis discriminate between SSc patients with and without ILD, seemingly outperforming conventional the ppTLC assessment. The relationship between radiomics and PFTs appeared predominantly in subjects with ILD while was undetectable in those without ILD.