Assist clinicians interpreting chest X-ray studies by detecting up to 124 findings
Assist clinicians interpreting chest X-ray studies by detecting up to 124 findings
Non-contrast CT (DICOM)
DICOM, HL7
When fixed at a sensitivity comparable to radiology reports (87.2%), the AI produced critical miss rates of 2.2%, compared to 1.1% for radiologists. At higher sensitivities (≥95.4%), the AI reduced critical misses to ≤1.1%, equaling or outperforming radiologists in detecting critical conditions without increasing the risk of missed diagnoses
In a study of 1961 chest radiographs, the system achieved a specificity of 24.5%, 47.1%, and 52.7% at sensitivities of 99.9%, 99.0%, and 98.0%, respectively. This allowed the AI to correctly exclude pathology in 19.6% of cases at the lowest sensitivity level, thus offering significant time savings for radiologists without increasing the risk of critical misses
In a study involving 2972 chest radiographs, the product contributed to significant report changes in 3.1% of cases, while changes in patient management occurred in 1.4% of cases. Furthermore, additional imaging was recommended in 1.0% of cases due to AI findings
Out of 2972 cases, the AI model findings aligned with radiologist interpretations in 86.5% of cases, with significant findings affecting patient management in 1.4% of instances
Plesner, L.L., Müller, F.C., Brejnebøl, M.W., et al. (2024). Using AI to Identify Unremarkable Chest Radiographs for Automatic Reporting. Radiology. DOI: 10.1148/radiol.240272Jones, C. M., et al. (2021). Assessment of the effect of a comprehensive chest radiograph deep learning model on radiologist reports and patient outcomes: a real-world observational study. BMJ Open, 11, e052902.
