Related Experiment Video
Updated: Jul 30, 2025
07:53
Author Spotlight: Advancing 3D Modeling for Enhanced Diagnosis and Treatment of Pulmonary Nodules in Early-Stage Lung Cancer
Published on: October 13, 2023
1.5K
Deep learning to estimate lung disease mortality from chest radiographs
Jakob Weiss1,2,3,4, Vineet K Raghu1,4, Dennis Bontempi1,5
1Artificial Intelligence in Medicine (AIM) Program, Mass General Brigham, Harvard Medical School, Harvard Institutes of Medicine, 77 Avenue Louis Pasteur, Boston, MA, 02115, USA.
Nature Communications
|May 16, 2023
Summary
A new deep learning model, CXR Lung-Risk, predicts lung disease mortality risk from chest X-rays. This tool aids in identifying high-risk individuals for personalized prevention and treatment strategies.
Area of Science:
- Artificial Intelligence in Medicine
- Radiology and Medical Imaging
Background:
- Chronic lung diseases pose significant health challenges.
- Accurate prediction of severe outcomes remains limited.
- Chest X-rays are widely available diagnostic tools.
Purpose of the Study:
- To develop and validate a deep learning model (CXR Lung-Risk) for predicting lung disease mortality risk.
- To assess the model's performance using chest X-ray data.
- To evaluate the model's utility in improving risk stratification.
Main Methods:
- Development of a deep learning model, CXR Lung-Risk.
- Training on a large dataset of 147,497 chest X-ray images from 40,643 individuals.
- Validation in three independent cohorts totaling 15,976 individuals.
Main Results:
- CXR Lung-Risk demonstrated a significant graded association with lung disease mortality (HR up to 11.86).
- The model's predictions were adjusted for key risk factors like age and smoking.
- Integration of CXR Lung-Risk improved mortality risk estimation across all tested cohorts.
Conclusions:
- Deep learning applied to chest X-rays can effectively identify individuals at high risk of lung disease mortality.
- CXR Lung-Risk offers a promising tool for enhancing personalized prevention and treatment strategies.
- This approach leverages easily obtainable imaging data for improved patient outcomes.
