Deep learning for chest radiograph diagnosis: A retrospective comparison of the CheXNeXt algorithm to practicing radiologists

Key Points

• To develop and validate a deep learning algorithm for classifying abnormalities in chest radiographs, matching the performance of practicing radiologists.
• Developed a deep learning algorithm based on chest radiographs
• Conducted a retrospective comparison with radiologists' diagnoses
• Valued performance in classifying clinically important abnormalities
• Deep learning algorithm performed at a level comparable to radiologists in diagnosing abnormalities
• Potential to improve patient access to chest radiograph diagnostics
• Validation indicates robustness for future clinical implementation