Related Experiment Video
Updated: Dec 21, 2025
08:05
Lung CT Segmentation to Identify Consolidations and Ground Glass Areas for Quantitative Assesment of SARS-CoV Pneumonia
Published on: December 19, 2020
14.6K
Automated abnormality classification of chest radiographs using deep convolutional neural networks
Yu-Xing Tang1, You-Bao Tang1, Yifan Peng2
11Imaging Biomarkers and Computer-Aided Diagnosis Laboratory, Radiology and Imaging Sciences, National Institutes of Health Clinical Center, Bethesda, MD 20892 USA.
NPJ Digital Medicine
|May 22, 2020
Summary
Deep convolutional neural networks (CNNs) accurately differentiate normal and abnormal chest X-rays, improving radiology workflow. This AI tool aids in prioritizing radiograph interpretation and disease diagnosis.
Area of Science:
- Radiology
- Artificial Intelligence
- Medical Imaging
Background:
- Chest radiography is a vital diagnostic tool requiring efficient interpretation.
- Timely detection of abnormalities in chest X-rays is crucial for patient care and workflow management.
- Automated analysis of medical images can significantly enhance diagnostic speed and accuracy.
Purpose of the Study:
- To develop and evaluate deep convolutional neural networks (CNNs) for classifying normal versus abnormal chest radiographs.
- To assess the performance of CNN models in identifying specific conditions like lung opacity and pneumonia.
- To determine the generalizability and potential benefits of AI in radiology workflow prioritization.
Main Methods:
- Development and evaluation of various deep convolutional neural networks (CNNs).
- Training and testing models on datasets of frontal chest radiographs.
- Fine-tuning models pre-trained on adult patients for pediatric cases.
- Utilizing natural images for pre-training CNNs.
Main Results:
- A CNN model achieved an AUC of 0.9824 for normal vs. abnormal classification.
- The model demonstrated high accuracy (94.64%) and sensitivity (96.50%) in differentiating normal from abnormal radiographs.
- Excellent performance was observed for lung opacity (AUC 0.9804) and pneumonia classification (AUC 0.9851).
- The model showed strong generalizability on an external dataset (AUC 0.9444).
Conclusions:
- Deep CNNs can accurately and effectively differentiate normal from abnormal chest radiographs.
- AI-powered tools show potential for improving radiology workflow through work list triaging and reporting prioritization.
- The findings suggest significant benefits for patient care via faster and more reliable image interpretation.
