Late Breaking Abstract - Deep-learning Algorithm FibroNet to Identify Fibrotic Primary Sjögren’s Syndrome-associated Interstitial Lung Disease based on Chest HRCT

<bold>Background:</bold> The prognosis of fibrotic primary Sjogren’s syndrome associated interstitial lung disease (F-pSS-ILD) is worse than that of non-fibrotic (NF)-pSS-ILD. However, it is challenging to use chest High-resolution computed tomography (HRCT) to identify F-pSS-ILD in non-professional ILD centers. In this study, we developed a deep learning model for identifying F-pSS-ILD. <bold>Methods:</bold> A total of 996 HRCTs from 399 patients with pSS-ILD in ILD-China cohort were included in this study. A deep learning model named FibroNet was developed for identifying F-pSS-ILD. To assess the effectiveness of this model, we compared it with Lung Graph (Sun H, et al. J Imaging Inform Med. 2024 Feb;37(1):268-279.) and four groups of doctors (three doctors per group). We also generate a heatmap, visualizing the important regions that led to the decision of our deep learning model. <bold>Results:</bold> FibroNet was better than Lung Graph, pulmonologists and radiologists who without ILD experience. The diagnostic performance was similar to that of pulmonologists with ILD experience, and slightly inferior to that of radiologists with ILD experience. <bold>Conclusions:</bold> We developed a deep learning model (FibroNet) that uses chest HRCT images to identify F-pSS-ILD, which can identify fibrotic types, including not only UIP but also fNSIP. <fig><object-id>erj;64/suppl_68/PA3415/F1</object-id><object-id>F1</object-id><object-id>F1</object-id><graphic></graphic></fig>