A deep learning model to identify fluid overload status in critically ill patients based on chest X-ray images.

Researchers

Journal

Modalities

Models

Abstract

Recent studies have highlighted adverse outcomes of fluid overload in critically ill patients. Therefore, early recognition is essential for the management of these patients.Our purpose is to propose a deep learning (DL) model to explore noninvasive chest X-ray (CXR) image information associated with fluid overload status.We collected study data from the Medical Information Mart for Intensive Care IV (MIMIC-IV, v1.0) and MIMIC Chest X-Ray (v2.0.0) databases for modeling, and from our hospital for testing. Extravascular lung water index (ELWI) > 10 mL/kg and global end diastolic volume index (GEDI) > 700 mL/m2 were used as threshold values for fluid overload status. A DL model with a transfer learning strategy was proposed to predict fluid overload status through CXR images in comparison with clinical and semantic label models.Whether in the primary cohort or test cohort, the DL models showed relatively strong performance for predicting the ELWI (AUROC: 0.896, 95% CI 0.819-0.972 and 0.718, 0.594-0.822, respectively) and GEDI status (AUROC: 0.814, 95% CI 0.699-0.930 and 0.649, 0.510-0.787, respectively), which were better than clinical and semantic label models. Additionally, a visualization technique to determine the important areas of features in the input images was adopted.As CXR is routinely used in the intensive care unit, a simple, fast, low-cost, and noninvasive DL model can be regarded as an effective supplementary tool for identifying fluid overload, and should be widely adopted in clinical applications, especially when invasive hemodynamic monitoring is not available.