Big data insights in the ILD population through AI-driven analytics

Le résumé fourni par la source

Rationale: ILD is a rare progressive disease with a huge burden. We aim to utilize AI model for pulmonary function testing (ArtiQ.PFT) to leverage ILD related data insights. Understanding the ILD population in a specific region would enable better allocation of resources for healthcare industries. Methods: ArtiQ.PFT was applied to anonymized PFT data from 87205 patients from UZ Leuven (the reference center) and 7 referring centers throughout Belgium. We compared the percentages of ILD-predicted patients, patients with suggestive restrictive spirometry, patients with restrictive volume, and patients with normal spirometry but reduced DLCO. Results: AI predicted 12366 patients (14%) to have ILD, leading to an overall higher patient percentage at the reference center (26%) compared to the referring centers (8%,SD 3%). The difference was driven by a high percentage of patients with restrictive volume (22% vs. 12%,SD 2%) and a high percentage of patients with normal spirometry but reduced DLCO (27% vs. 14%,SD 2%). Surprisingly, only 15% of patients at the reference center exhibited suggestive restrictive spirometry, compared to the referring centers (14%,SD 4%). Conclusions: ArtiQ.PFT was used to identify specific variances in the predicted-ILD profiles using a large sample across Belgium. Future research is still needed to explain if observed differences are driven by bad quality spirometry, different diagnosis strategy or different disease population. erj;64/suppl_68/PA2466/F1 F1 F1

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.