Purpose
Chest X-ray (CXR) is the most commonly performed radiological exam in the world. In 2006, it was estimated that 129 million CXR images were acquired in the United States alone [1]. CXR is often the first imaging study acquired and remains central to screening, diagnosis, and management of a broad range of conditions. It is still widely-used because of its cost-effectiveness and low radiation dose. The interpretation of CXR can however be challenging due to the superimposition of anatomical structures, difference in the image quality,...
Methods and materials
Thoracic images of patients who underwent CXR and had a thoracic CT scan within 72 hours at Cochin university hospital (AP-HP, Paris, France) were collected over a 10-year period (2010-2020). A senior radiologist specialized in thoracic imaging annotated the CXR images for five main anomalies (pneumothorax, pleural effusion, mediastino-hilar mass, nodule, and consolidation), with the corresponding CT scan as the standard of reference.. Each abnormality was thus classified by the same chest radiologist into two different categories: detectable on the CXR or only detectable on...
Results
Five hundred patients were included in the study with mean age years 53.5 ± 18.6, of which 261 (52.2%) were females. Of the 500 examinations, 243 examinations had one of the five main abnormalities detectable on the CXR whereas 257 showed no abnormality. Overall, there were 276 examinations with at least one of the five main abnormalities seen on CT.
The AI algorithm had a sensitivity of 88.1% for the classification of CXR as normal or abnormal based on the CT findings and a specificity...
Conclusion
These preliminary results show that the AI algorithm has higher sensitivity than all readers and equivalent specificity to experts for detecting CXR abnormalities. This study shows that the use of the AI has the potential to decrease diagnostic errors.
Future studies should be carried out to analyze the interaction between the AI algorithm and the readers to determine if the use of the software can improve their performance to detect chest abnormalities.
Personal information and conflict of interest
S. Bennani:
Consultant: Gleamer
N-e. Regnard:
Founder: Gleamer
L. Lassalle:
Consultant: Gleamer
T. Nguyen:
Consultant: Gleamer
C. Malandrin:
Other: Gleamer
H. Koulakian:
Other: Gleamer
P. Khafagy:
Other: Gleamer
G. Chassagnon:
Other: Gleamer
M-P. Revel:
Nothing to disclose
References
Mettler Jr, F. A., Bhargavan, M., Faulkner, K., Gilley, D. B., Gray, J. E., Ibbott, G. S., ... & Yoshizumi, T. T. (2009). Radiologic and nuclear medicine studies in the United States and worldwide: frequency, radiation dose, and comparison with other radiation sources—1950–2007. Radiology, 253(2), 520-531.
Çallı, E., Sogancioglu, E., van Ginneken, B., van Leeuwen, K. G., & Murphy, K. (2021). Deep learning for chest X-ray analysis: A survey. Medical Image Analysis, 72, 102125.
Duron, L., Ducarouge, A., Gillibert, A., Lainé, J., Allouche, C., Cherel,...