A deep learning model enables accurate prediction and quantification of pulmonary edema from chest X-rays

A study assessed the severity of pulmonary edema using quantitative measures obtained from chest radiography. The researchers utilized a deep learning approach to predict the severity of pulmonary edema based on the extravascular lung water index (EVLWI), which is an invasive measure extracted from transpulmonary thermodilution. The study included 471 X-rays from 431 patients who underwent both chest radiography and EVLWI measurement. The deep learning models achieved high accuracy, with an AUROC (area under the receiver operating characteristic curve) ranging from 0.97 to 0.99 and an MCC (Mathews correlation coefficient) between 0.86 and 0.92 in multiclass models. This study indicates that deep learning can effectively quantify pulmonary edema with accuracy.