Background: Interstitial lung disease (ILD) is a common and serious complication of connective tissue diseases (CTD), presenting various tomographic patterns. However, the extent and severity of these patterns can vary significantly depending on the CTD and autoantibodies (AAb) present in patients (pts). Understanding these variations is crucial for improving management and treatment. This single-center pilot study used artificial intelligence in the segmentation and classification process to analyze tomographic patterns in non-smoking patients with ILD associated with different CTDs, to explore the correlation between different AAb positivity and volumetric extension (VE) of lung abnormalities detected by high-resolution CT (HRCT), such as Ground Glass (GG), Honeycombing (HC), Bronchiectasis (BcE), and vascular quota (VQ). The accuracy of this approach was validated through ROC curve analysis, which demonstrated perfect discrimination for Ground Glass and Honeycombing patterns (AUC = 1.000), excellent accuracy for vascular assessment (AUC = 0.943, 95% CI 0.840-1.000), and good discrimination for bronchiectasis (AUC = 0.757, 95% CI 0.619-0.894).

Objectives: The primary objective of this study was to investigate the correlation between AAb profiles and the VE of ILD tomographic patterns in non-smoking patients with CTDs.

Methods: 158 HRCT scans of patients with CTD were examined (rheumatoid arthritis: 24, UCTD: 23, myositis: 8, SLE: 3, overlap: 10, Sjögren’s syndrome: 9, systemic sclerosis: 79), mean age 51.2±12.7 years, and disease duration 12.2±9.1 years. Different tomographic patterns were examined using 3D Slicer ver. 5.7, particularly GG, HC, BcE, and VQ (Figure 1). ILD was diagnosed in 97 patients. The segmentation of various lung patterns was performed through an artificial intelligence-guided process. This program was used considering its efficacy studies in patients affected by SarsCov2. Quality control of segmentation was performed by an operator experienced in interstitial lung disease analysis. The CT scans examined were performed with the same instrument to exclude possible analysis bias by the program, and CT scans showing infectious or neoplastic processes were excluded. None of the patients had previous antifibrotic therapy with nintedanib or pirfenidone. We analyzed the correlation between AAb profiles and tomographic patterns. ROC curve analysis confirmed the high accuracy of the artificial intelligence-guided segmentation process, particularly for Ground Glass and Honeycombing patterns (both with AUC = 1.000, p < 0.001) and vascular assessment (AUC = 0.943, p = 0.001). While bronchiectasis segmentation demonstrated good accuracy (AUC = 0.757, p = 0.051), it exhibited greater variability.

Results: 23% of patients with CTD-ILD had only ANA+, 15% anti-Scl-70+, 18% CENP-B+, 6% rheumatoid factor (RF)+, 3% ACPA+, 8% anti-SSA+, 2% had other anti-ENA specificities, 26% had 2 or more AAb positivity (M-AAb+). Patients with anti-Scl-70+ showed the highest VE of BcE 132.2±387.4 cm3 (p=0.049). The VE of GG was significantly higher in patients with anti-Scl-70+ (667.8±313.6 cm3) compared to patients with only ANA+ (474.3±371.7cm3, p=0.043), CENP-B+ (187.2±200.6 cm3, p=0.0001), RF+ (339.6±208.6 cm3, p=0.021) and anti-SS-A+ (394.7±225.6cm3, p=0.033); it was also significantly higher in patients with only ANA+ (474.3±371.7cm3, p=0.002) and those with M-AAb+ (486.7±280.2 cm3, p=0.001 OR: 2.7) compared to patients with CENP-B+ (187.2±200.6). The VE of HC was significantly higher in patients with only ANA+ (80.1±61.4cm3), anti-Scl-70+ (97.7±60.4cm3) and M-AAb+ (78.9±60.5 cm3) compared to patients with CENP-B+ (30.6±43 cm3) (Table 1). Linear regression analysis showed that greater volumetric extension of HC is associated with the presence of anti-Scl-70 (OR: 2.1; p=0.035); while greater extension of GG is associated with the presence of anti-Scl-70 (OR: 2.7; p=0.008) and ACPA (OR=1.9; p=0.05).

Conclusion: The study showed significant differences in the VE of CTD-ILD tomographic patterns based on AAb profiles. In particular, it confirms, even through the use of software equipped with artificial intelligence, that the most severe ILD patterns are associated with positivity for anti-Scl-70 (for all tomographic patterns, such as BcE, GG, and HC) and the presence of M-AAb+ (for GG and HC patterns). It also shows that anti-Scl-70 is more associated with the greater presence of fibrosing pattern, as well as the presence of M-AAb+, and that ACPA positivity is associated with a higher probability of greater GG extension. Therefore, the clinical importance of particular antibody subsets in predicting more extensive forms of ILD is emphasized, to personalize the clinical and therapeutic management of patients with CTD. However, further large-scale, prospective studies are needed to confirm these observations and potentially revolutionize the therapeutic approach to ILD in CTD.

REFERENCES: NIL.

Acknowledgements: NIL.

Disclosure of Interests: None declared.

© The Authors 2025. This abstract is an open access article published in Annals of Rheumatic Diseases under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/ ). Neither EULAR nor the publisher make any representation as to the accuracy of the content. The authors are solely responsible for the content in their abstract including accuracy of the facts, statements, results, conclusion, citing resources etc.