RT Journal Article SR Electronic T1 Machine learning (ML) classifier models in ARDS to identify patients with a low percentage of potentially recruitable lung JF European Respiratory Journal JO Eur Respir J FD European Respiratory Society SP 2383 DO 10.1183/13993003.congress-2022.2383 VO 60 IS suppl 66 A1 F Pennati A1 S Coppola A1 A Aliverti A1 D Chiumello YR 2022 UL http://erj.ersjournals.com/content/60/suppl_66/2383.abstract AB In ARDS, the knowledge of the percentage of potentially recruitable lung, i.e. the proportion of the lung in which aeration is restored from 5 to 45 cmH2O of airway pressures, is important to establish the therapeutic efficacy of PEEP (Gattinoni, NEJM 2006). The accurate measure of the potentially recruitable lung is based on the acquisition of CT scans at both end-inspiration and end-expiration, but the X ray exposure must be carefully evaluated.The present work aimed to build ML models able to classify patients with a low percentage of potentially recruitable lung (<10%), using readily available clinical data and/or end-expiratory CT scan only, to reduce radiation exposure.Data from 221 patients with ARDS, who underwent whole-lung CT at airway pressures of 5 and 45 cmH2O were retrospectively collected. A random forest algorithm with a 10-fold cross validation was used to build three classification models based on measures of 1) clinical data (anthropometric, lung mechanics and gas exchange measures), 2) end-expiratory CT and 3) clinical and end-expiratory CT combined. Data were randomly split into a train/validation set (80%) and a test set (20%), to evaluate the performance of the model on unseen data.The areas under the ROC curve of the three models in the test set was 0.67, 0.71, 0.81, respectively. Accordingly, the sensitivities were 0.80, 0.85, 0.84, while the specificities were 0.53, 0.58, 0.79.ML provided a reliable tool to identify patients with a low percentage of potentially recruitable lung, thus it may provide clinical decision support in managing ARDS patients, preventing ventilator-induced lung injury while reducing radiation exposureFootnotesCite this article as Eur Respir J 2022; 60: Suppl. 66, 2383.This article was presented at the 2022 ERS International Congress, in session “-”.This is an ERS International Congress abstract. No full-text version is available. Further material to accompany this abstract may be available at www.ers-education.org (ERS member access only).